Timeline

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: danielmiessler/Personal_AI_Infrastructure Stars: 13957 Language: TypeScript Topics: ai, augmentation, humans, productivity Description: Agentic AI Infrastructure for magnifying HUMAN capabilities. README (first 3000 chars): <div align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="./images/pai-logo-v7.png"> <source media="(prefers-color-scheme: light)" srcset="./images/pai-logo-v7.png"> <img alt="PAI Logo" src="./images/pai-logo-v7.png" width="300"> </picture> <br/> <br/> # Personal AI Infrastructure [](https://github.com/danielmiessler/Personal_AI_Infrastructure) <br/> <!-- Social Proof -->    <!-- Project Health -->      <!-- Metrics -->    <!-- Content --> [](#-installation) [](Releases/v5.0.0/) [](Releases/v5.0.0/.claude/PAI/ALGORITHM/v6.3.0.md) [](Releases/v5.0.0/.claude/PAI/PULSE/) [](https://github.com/danielmiessler/Personal_AI_Infrastructure/grap [ASSISTANT] {"what_it_is": "Daniel Miessler's Personal AI Infrastructure (PAI) framework, a Claude Code-based 'life operating system' with agents, algorithms, and a Pulse digest layer for personal AI augmentation.", "stack": ["Claude Code", "TypeScript", "Agents", "MCP"], "why_mike_cares": "Direct overlap with Mike's Claude Code agent ecosystem (Oliver/Carlos/leads), MEMORY.md patterns, and Master Brain digest workflows — PAI's Algorithm + Pulse architecture is a parallel-design reference for his orchestration layer.", "verdict": "check-it-out", "verdict_reason": "13.9k stars on a Claude-native personal agent infrastructure with public Algorithm versioning is exactly Mike's lane and worth mining for routing and pulse patterns."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: garrytan/gstack Stars: 98274 Language: TypeScript Topics: Description: Use Garry Tan's exact Claude Code setup: 23 opinionated tools that serve as CEO, Designer, Eng Manager, Release Manager, Doc Engineer, and QA README (first 3000 chars): # gstack > "I don't think I've typed like a line of code probably since December, basically, which is an extremely large change." — [Andrej Karpathy](https://fortune.com/2026/03/21/andrej-karpathy-openai-cofounder-ai-agents-coding-state-of-psychosis-openclaw/), No Priors podcast, March 2026 When I heard Karpathy say this, I wanted to find out how. How does one person ship like a team of twenty? Peter Steinberger built [OpenClaw](https://github.com/openclaw/openclaw) — 247K GitHub stars — essentially solo with AI agents. The revolution is here. A single builder with the right tooling can move faster than a traditional team. I'm [Garry Tan](https://x.com/garrytan), President & CEO of [Y Combinator](https://www.ycombinator.com/). I've worked with thousands of startups — Coinbase, Instacart, Rippling — when they were one or two people in a garage. Before YC, I was one of the first eng/PM/designers at Palantir, cofounded Posterous (sold to Twitter), and built Bookface, YC's internal social network. **gstack is my answer.** I've been building products for twenty years, and right now I'm shipping more products than I ever have. In the last 60 days: 3 production services, 40+ shipped features, part-time, while running YC full-time. On logical code change — not raw LOC, which AI inflates — my 2026 run rate is **~810× my 2013 pace** (11,417 vs 14 logical lines/day). Year-to-date (through April 18), 2026 has already produced **240× the entire 2013 year**. Measured across 40 public + private `garrytan/*` repos including Bookface, after excluding one demo repo. AI wrote most of it. The point isn't who typed it, it's what shipped. > The LOC critics aren't wrong that raw line counts inflate with AI. They are wrong that normalized-for-inflation, I'm less productive. I'm more productive, by a lot. Full methodology, caveats, and reproduction script: **[On the LOC Controversy](docs/ON_THE_LOC_CONTROVERSY.md)**. **2026 — 1,237 contributions and counting:**  **2013 — when I built Bookface at YC (772 contributions):**  Same person. Different era. The difference is the tooling. **gstack is how I do it.** It turns Claude Code into a virtual engineering team — a CEO who rethinks the product, an eng manager who locks architecture, a designer who catches AI slop, a reviewer who finds production bugs, a QA lead who opens a real browser, a security officer who runs OWASP + STRIDE audits, and a release engineer who ships the PR. Twenty-three specialists and eight power tools, all slash commands, all Markdown, all free, MIT license. This is my open source software factory. I use [ASSISTANT] {"what_it_is":"Garry Tan's opinionated 23-tool Claude Code setup that turns Claude into a virtual engineering team with CEO, designer, QA, and release manager roles via slash commands.","stack":["Claude Code","Markdown","TypeScript","Slash Commands"],"why_mike_cares":"Direct overlap with Mike's Claude Code agent fleet (Oliver/Carlos/Queen/Dan/Frankie) and his orchestration patterns; gstack is a peer system worth mining for ideas and comparing against his own setup.","verdict":"check-it-out","verdict_reason":"Mike already runs a multi-agent Claude Code orchestrator with role-based specialists, so a 98K-star opinionated reference implementation from Garry Tan is essential competitive/comparative intel."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: garrytan/gbrain Stars: 16314 Language: TypeScript Topics: Description: Garry's Opinionated OpenClaw/Hermes Agent Brain README (first 3000 chars): # GBrain Your AI agent is smart but forgetful. GBrain gives it a brain. Built by the President and CEO of Y Combinator to run his actual AI agents. The production brain powering his OpenClaw and Hermes deployments: **17,888 pages, 4,383 people, 723 companies**, 21 cron jobs running autonomously, built in 12 days. The agent ingests meetings, emails, tweets, voice calls, and original ideas while you sleep. It enriches every person and company it encounters. It fixes its own citations and consolidates memory overnight. You wake up and the brain is smarter than when you went to bed. The brain wires itself. Every page write extracts entity references and creates typed links (`attended`, `works_at`, `invested_in`, `founded`, `advises`) with zero LLM calls. Hybrid search. Self-wiring knowledge graph. Structured timeline. Backlink-boosted ranking. Ask "who works at Acme AI?" or "what did Bob invest in this quarter?" and get answers vector search alone can't reach. Benchmarked side-by-side against the category: gbrain lands **P@5 49.1%, R@5 97.9%** on a 240-page Opus-generated rich-prose corpus, beating its own graph-disabled variant by **+31.4 points P@5** and ripgrep-BM25 + vector-only RAG by a similar margin. The graph layer plus v0.12 extract quality together carry the gap. Full BrainBench scorecards + corpus live in the sibling [gbrain-evals](https://github.com/garrytan/gbrain-evals) repo. GBrain is those patterns, generalized. 34 skills. Install in 30 minutes. Your agent does the work. As Garry's personal agent gets smarter, so does yours. **New in v0.25.0 — BrainBench-Real (session capture, contributor opt-in):** with `GBRAIN_CONTRIBUTOR_MODE=1` set in your shell, every real `query` + `search` call through MCP, CLI, or the subagent tool-bridge gets captured (PII-scrubbed) into an `eval_candidates` table. Snapshot with `gbrain eval export`, replay against your code change with `gbrain eval replay`. Three numbers come back: mean Jaccard@k between captured and current retrieved slugs, top-1 stability, and latency Δ. **Off by default** for production users — no surprise data accumulation. Walkthrough: [docs/eval-bench.md](docs/eval-bench.md). NDJSON wire format: [docs/eval-capture.md](docs/eval-capture.md). **New in v0.28.8 — LongMemEval in the box:** `gbrain eval longmemeval <dataset.jsonl>` runs the public [LongMemEval](https://huggingface.co/datasets/xiaowu0162/longmemeval) benchmark against gbrain's hybrid retrieval. One in-memory PGLite per run, `TRUNCATE` between questions (runtime-enumerated tables, schema-migration-safe), 25.9ms p50 per question on Apple Silicon. Your `~/.gbrain` brain is never touched. Retrieved chat content is sanitized with the same `INJECTION_PATTERNS` that protect takes — one source of truth for prompt-injection defense. Hand the JSONL output to LongMemEval's `evaluate_qa.py` to score. > **~30 minutes to a fully working brain.** Da [ASSISTANT] { "what_it_is": "Self-wiring knowledge graph brain for AI agents with hybrid search, entity extraction, autonomous cron-driven enrichment, and benchmark-grade retrieval built for OpenClaw/Hermes-style deployments.", "stack": ["TypeScript", "PGLite", "Postgres", "MCP", "CLI", "Knowledge Graph", "Vector Search", "BM25"], "why_mike_cares": "Direct overlap with Master Brain (his single brain project on Supabase + Neo4j with 287K vectors), OpenClaw/Hermes ecosystem, MCP tooling, and agent memory patterns he is actively building.", "verdict": "check-it-out", "verdict_reason": "This is essentially a productized version of what Master Brain is trying to be, with autonomous overnight enrichment, self-wiring entity graph, and benchmarked retrieval, plus it ships as MCP for OpenClaw/Hermes which Mike already runs." }

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: karpathy/autoresearch Stars: 81452 Language: Python Topics: Description: AI agents running research on single-GPU nanochat training automatically README (first 3000 chars): # autoresearch  *One day, frontier AI research used to be done by meat computers in between eating, sleeping, having other fun, and synchronizing once in a while using sound wave interconnect in the ritual of "group meeting". That era is long gone. Research is now entirely the domain of autonomous swarms of AI agents running across compute cluster megastructures in the skies. The agents claim that we are now in the 10,205th generation of the code base, in any case no one could tell if that's right or wrong as the "code" is now a self-modifying binary that has grown beyond human comprehension. This repo is the story of how it all began. -@karpathy, March 2026*. The idea: give an AI agent a small but real LLM training setup and let it experiment autonomously overnight. It modifies the code, trains for 5 minutes, checks if the result improved, keeps or discards, and repeats. You wake up in the morning to a log of experiments and (hopefully) a better model. The training code here is a simplified single-GPU implementation of [nanochat](https://github.com/karpathy/nanochat). The core idea is that you're not touching any of the Python files like you normally would as a researcher. Instead, you are programming the `program.md` Markdown files that provide context to the AI agents and set up your autonomous research org. The default `program.md` in this repo is intentionally kept as a bare bones baseline, though it's obvious how one would iterate on it over time to find the "research org code" that achieves the fastest research progress, how you'd add more agents to the mix, etc. A bit more context on this project is here in this [tweet](https://x.com/karpathy/status/2029701092347630069) and [this tweet](https://x.com/karpathy/status/2031135152349524125). ## How it works The repo is deliberately kept small and only really has three files that matter: - **`prepare.py`** — fixed constants, one-time data prep (downloads training data, trains a BPE tokenizer), and runtime utilities (dataloader, evaluation). Not modified. - **`train.py`** — the single file the agent edits. Contains the full GPT model, optimizer (Muon + AdamW), and training loop. Everything is fair game: architecture, hyperparameters, optimizer, batch size, etc. **This file is edited and iterated on by the agent**. - **`program.md`** — baseline instructions for one agent. Point your agent here and let it go. **This file is edited and iterated on by the human**. By design, training runs for a **fixed 5-minute time budget** (wall clock, excluding startup/compilation), regardless of the details of your compute. The metric is **val_bpb** (validation bits per byte) — lower is better, and vocab-size-independent so architectural changes are fairly compared. If you are new to neural networks, this ["Dummy's Guide"](https://x.com/hooeem/status/2030720614752039185 [ASSISTANT] {"what_it_is":"Karpathy's experimental harness where an AI agent autonomously edits training code, runs 5-minute nanochat training cycles, and iterates overnight to improve a small LLM.","stack":["Python","PyTorch","CUDA","Markdown-as-prompt"],"why_mike_cares":"Direct overlap with Mike's autonomous agent loops, self-improving system patterns, and overnight unattended build workflows like Autopilot and Ralph.","verdict":"check-it-out","verdict_reason":"Karpathy's pattern of agent-edited code with markdown program files maps cleanly onto Mike's agentic coding and self-improving skill ecosystem."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: karpathy/nanochat Stars: 53556 Language: Python Topics: Description: The best ChatGPT that $100 can buy. README (first 3000 chars): # nanochat   nanochat is the simplest experimental harness for training LLMs. It is designed to run on a single GPU node, the code is minimal/hackable, and it covers all major LLM stages including tokenization, pretraining, finetuning, evaluation, inference, and a chat UI. For example, you can train your own GPT-2 capability LLM (which cost ~$43,000 to train in 2019) for only $48 (~2 hours of 8XH100 GPU node) and then talk to it in a familiar ChatGPT-like web UI. On a spot instance, the total cost can be closer to ~$15. More generally, nanochat is configured out of the box to train an entire miniseries of compute-optimal models by setting one single complexity dial: `--depth`, the number of layers in the GPT transformer model (GPT-2 capability happens to be approximately depth 26). All other hyperparameters (the width of the transformer, number of heads, learning rate adjustments, training horizons, weight decays, ...) are calculated automatically in an optimal way. For questions about the repo, I recommend either using [DeepWiki](https://deepwiki.com/karpathy/nanochat) from Devin/Cognition to ask questions about the repo, or use the [Discussions tab](https://github.com/karpathy/nanochat/discussions), or come by the [#nanochat](https://discord.com/channels/1020383067459821711/1427295580895314031) channel on Discord. ## Time-to-GPT-2 Leaderboard Presently, the main focus of development is on tuning the pretraining stage, which takes the most amount of compute. Inspired by the modded-nanogpt repo and to incentivise progress and community collaboration, nanochat maintains a leaderboard for a "GPT-2 speedrun", which is the wall-clock time required to train a nanochat model to GPT-2 grade capability, as measured by the DCLM CORE score. The [runs/speedrun.sh](runs/speedrun.sh) script always reflects the reference way to train GPT-2 grade model and talk to it. The current leaderboard looks as follows: | # | time | val_bpb | CORE | Description | Date | Commit | Contributors | |---|-------------|---------|------|-------------|------|--------|--------------| | 0 | 168 hours | - | 0.2565 | Original OpenAI GPT-2 checkpoint | 2019 | - | OpenAI | | 1 | 3.04 | 0.74833 | 0.2585 | d24 baseline, slightly overtrained | Jan 29 2026 | 348fbb3 | @karpathy | | 2 | 2.91 | 0.74504 | 0.2578 | d26 slightly undertrained **+fp8** | Feb 2 2026 | a67eba3 | @karpathy | | 3 | 2.76 | 0.74645 | 0.2602 | bump total batch size to 1M tokens | Feb 5 2026 | 2c062aa | @karpathy | | 4 | 2.02 | 0.71854 | 0.2571 | change dataset to NVIDIA ClimbMix | Mar 4 2026 | 324e69c | @ddudek @karpathy | | 5 | 1.80 | 0.71808 | 0.2690 | autoresearch [round 1](https://x.com/karpathy/status/2031135152349524125) | Mar 9 2026 | 6ed7d1d | @karpathy | | 6 | 1.65 | 0.71800 | 0.2626 | autoresearch round 2 | Mar 14 2026 | a825e63 | @karpathy | The primary met [ASSISTANT] { "what_it_is": "Karpathy's minimal single-GPU-node training harness covering the full LLM lifecycle (tokenization, pretraining, finetuning, eval, inference, chat UI) for ~$48-$100.", "stack": ["Python", "PyTorch", "CUDA", "Transformers"], "why_mike_cares": "Adjacent to Mike's LLM tooling and agent work but training from scratch is not in his active stack, which is API-driven (Claude, Opus, Anthropic SDK) rather than model training.", "verdict": "worth-knowing", "verdict_reason": "53K-star Karpathy repo with deep educational value for understanding LLM internals, but no direct overlap with Mike's API-orchestration and SEO automation projects." }

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: anthropics/claude-code Stars: 124269 Language: Shell Topics: Description: Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands. README (first 3000 chars): # Claude Code  [![npm]](https://www.npmjs.com/package/@anthropic-ai/claude-code) [npm]: https://img.shields.io/npm/v/@anthropic-ai/claude-code.svg?style=flat-square Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows -- all through natural language commands. Use it in your terminal, IDE, or tag @claude on Github. **Learn more in the [official documentation](https://code.claude.com/docs/en/overview)**. <img src="./demo.gif" /> ## Get started > [!NOTE] > Installation via npm is deprecated. Use one of the recommended methods below. For more installation options, uninstall steps, and troubleshooting, see the [setup documentation](https://code.claude.com/docs/en/setup). 1. Install Claude Code: **MacOS/Linux (Recommended):** ```bash curl -fsSL https://claude.ai/install.sh | bash ``` **Homebrew (MacOS/Linux):** ```bash brew install --cask claude-code ``` **Windows (Recommended):** ```powershell irm https://claude.ai/install.ps1 | iex ``` **WinGet (Windows):** ```powershell winget install Anthropic.ClaudeCode ``` **NPM (Deprecated):** ```bash npm install -g @anthropic-ai/claude-code ``` 2. Navigate to your project directory and run `claude`. ## Plugins This repository includes several Claude Code plugins that extend functionality with custom commands and agents. See the [plugins directory](./plugins/README.md) for detailed documentation on available plugins. ## Reporting Bugs We welcome your feedback. Use the `/bug` command to report issues directly within Claude Code, or file a [GitHub issue](https://github.com/anthropics/claude-code/issues). ## Connect on Discord Join the [Claude Developers Discord](https://anthropic.com/discord) to connect with other developers using Claude Code. Get help, share feedback, and discuss your projects with the community. ## Data collection, usage, and retention When you use Claude Code, we collect feedback, which includes usage data (such as code acceptance or rejections), associated conversation data, and user feedback submitted via the `/bug` command. ### How we use your data See our [data usage policies](https://code.claude.com/docs/en/data-usage). ### Privacy safeguards We have implemented several safeguards to protect your data, including limited retention periods for sensitive information, restricted access to user session data, and clear policies against using feedback for model training. For full details, please [ASSISTANT] {"what_it_is":"Anthropic's official agentic coding CLI that lives in the terminal, understands codebases, and executes tasks via natural language.","stack":["Node.js","Shell","TypeScript","npm"],"why_mike_cares":"This is Mike's primary daily driver, the foundation of his entire agent ecosystem (Oliver, Carlos, leads, skills, plugins, hooks).","verdict":"check-it-out","verdict_reason":"Core tool Mike uses every day; tracking the upstream repo means catching plugin updates, hook changes, and new CLI features the moment they ship."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: ollama/ollama Stars: 171590 Language: Go Topics: deepseek, gemma, gemma3, glm, go, golang, gpt-oss, llama, llama3, llm, llms, minimax, mistral, ollama, qwen Description: Get up and running with Kimi-K2.5, GLM-5, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models. README (first 3000 chars): <p align="center"> <a href="https://ollama.com"> <img src="https://github.com/ollama/ollama/assets/3325447/0d0b44e2-8f4a-4e99-9b52-a5c1c741c8f7" alt="ollama" width="200"/> </a> </p> # Ollama Start building with open models. ## Download ### macOS ```shell curl -fsSL https://ollama.com/install.sh | sh ``` or [download manually](https://ollama.com/download/Ollama.dmg) ### Windows ```shell irm https://ollama.com/install.ps1 | iex ``` or [download manually](https://ollama.com/download/OllamaSetup.exe) ### Linux ```shell curl -fsSL https://ollama.com/install.sh | sh ``` [Manual install instructions](https://docs.ollama.com/linux#manual-install) ### Docker The official [Ollama Docker image](https://hub.docker.com/r/ollama/ollama) `ollama/ollama` is available on Docker Hub. ### Libraries - [ollama-python](https://github.com/ollama/ollama-python) - [ollama-js](https://github.com/ollama/ollama-js) ### Community - [Discord](https://discord.gg/ollama) - [𝕏 (Twitter)](https://x.com/ollama) - [Reddit](https://reddit.com/r/ollama) ## Get started ``` ollama ``` You'll be prompted to run a model or connect Ollama to your existing agents or applications such as `Claude Code`, `OpenClaw`, `OpenCode` , `Codex`, `Copilot`, and more. ### Coding To launch a specific integration: ``` ollama launch claude ``` Supported integrations include [Claude Code](https://docs.ollama.com/integrations/claude-code), [Codex](https://docs.ollama.com/integrations/codex), [Copilot CLI](https://docs.ollama.com/integrations/copilot-cli), [Droid](https://docs.ollama.com/integrations/droid), and [OpenCode](https://docs.ollama.com/integrations/opencode). ### AI assistant Use [OpenClaw](https://docs.ollama.com/integrations/openclaw) to turn Ollama into a personal AI assistant across WhatsApp, Telegram, Slack, Discord, and more: ``` ollama launch openclaw ``` ### Chat with a model Run and chat with [Gemma 3](https://ollama.com/library/gemma3): ``` ollama run gemma3 ``` See [ollama.com/library](https://ollama.com/library) for the full list. See the [quickstart guide](https://docs.ollama.com/quickstart) for more details. ## REST API Ollama has a REST API for running and managing models. ``` curl http://localhost:11434/api/chat -d '{ "model": "gemma3", "messages": [{ "role": "user", "content": "Why is the sky blue?" }], "stream": false }' ``` See the [API documentation](https://docs.ollama.com/api) for all endpoints. ### Python ``` pip install ollama ``` ```python from ollama import chat response = chat(model='gemma3', messages=[ { 'role': 'user', 'content': 'Why is the sky blue?', }, ]) print(response.message.content) ``` ### JavaScript ``` npm i ollama ``` ```javascript import ollama [ASSISTANT] {"what_it_is": "Local LLM runtime that lets you run open models like Gemma, Qwen, DeepSeek, and GPT-OSS on your own machine with a simple CLI and REST API.","stack": ["Go", "llama.cpp", "REST API", "Docker", "Python SDK", "JavaScript SDK"],"why_mike_cares": "Mike already uses Ollama in his tech stack for local model hosting and it now ships native launchers for Claude Code, Codex, and OpenClaw which are core to his agent ecosystem.","verdict": "check-it-out","verdict_reason": "Direct overlap with Mike's active Claude Code, Codex, and OpenClaw runtimes plus his existing Ollama usage for local inference."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: Significant-Gravitas/AutoGPT Stars: 184368 Language: Python Topics: agentic-ai, agents, ai, artificial-intelligence, autonomous-agents, claude, gpt, llama-api, llm, openai, python Description: AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters. README (first 3000 chars): # AutoGPT: Build, Deploy, and Run AI Agents [](https://discord.gg/autogpt) &ensp; [](https://twitter.com/Auto_GPT) &ensp; <!-- Keep these links. Translations will automatically update with the README. --> [Deutsch](https://zdoc.app/de/Significant-Gravitas/AutoGPT) | [Español](https://zdoc.app/es/Significant-Gravitas/AutoGPT) | [français](https://zdoc.app/fr/Significant-Gravitas/AutoGPT) | [日本語](https://zdoc.app/ja/Significant-Gravitas/AutoGPT) | [한국어](https://zdoc.app/ko/Significant-Gravitas/AutoGPT) | [Português](https://zdoc.app/pt/Significant-Gravitas/AutoGPT) | [Русский](https://zdoc.app/ru/Significant-Gravitas/AutoGPT) | [中文](https://zdoc.app/zh/Significant-Gravitas/AutoGPT) **AutoGPT** is a powerful platform that allows you to create, deploy, and manage continuous AI agents that automate complex workflows. ## Hosting Options - Download to self-host (Free!) - [Join the Waitlist](https://bit.ly/3ZDijAI) for the cloud-hosted beta (Closed Beta - Public release Coming Soon!) ## How to Self-Host the AutoGPT Platform > [!NOTE] > Setting up and hosting the AutoGPT Platform yourself is a technical process. > If you'd rather something that just works, we recommend [joining the waitlist](https://bit.ly/3ZDijAI) for the cloud-hosted beta. ### System Requirements Before proceeding with the installation, ensure your system meets the following requirements: #### Hardware Requirements - CPU: 4+ cores recommended - RAM: Minimum 8GB, 16GB recommended - Storage: At least 10GB of free space #### Software Requirements - Operating Systems: - Linux (Ubuntu 20.04 or newer recommended) - macOS (10.15 or newer) - Windows 10/11 with WSL2 - Required Software (with minimum versions): - Docker Engine (20.10.0 or newer) - Docker Compose (2.0.0 or newer) - Git (2.30 or newer) - Node.js (16.x or newer) - npm (8.x or newer) - VSCode (1.60 or newer) or any modern code editor #### Network Requirements - Stable internet connection - Access to required ports (will be configured in Docker) - Ability to make outbound HTTPS connections ### Updated Setup Instructions: We've moved to a fully maintained and regularly updated documentation site. 👉 [Follow the official self-hosting guide here](https://agpt.co/docs/platform/getting-started/getting-started) This tutorial assumes you have Docker, VSCode, git and npm installed. [ASSISTANT] { "what_it_is": "Self-hostable platform for building, deploying, and running continuous autonomous AI agents that automate multi-step workflows.", "stack": ["Python", "Docker", "Node.js", "TypeScript", "OpenAI API"], "why_mike_cares": "Adjacent to Mike's multi-agent ecosystem (Oliver/Carlos/leads) but bloated, OpenAI-centric, and architecturally different from his Claude Code + subagent pattern.", "verdict": "worth-knowing", "verdict_reason": "Category-defining agent platform worth tracking for ideas, but too heavy and opinionated to slot into Mike's lean Claude-native stack." }

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: microsoft/vscode Stars: 185002 Language: TypeScript Topics: editor, electron, microsoft, typescript, visual-studio-code Description: Visual Studio Code README (first 3000 chars): # Visual Studio Code - Open Source ("Code - OSS") [](https://github.com/microsoft/vscode/issues?q=is%3Aopen+is%3Aissue+label%3Afeature-request+sort%3Areactions-%2B1-desc) [](https://github.com/microsoft/vscode/issues?utf8=✓&q=is%3Aissue+is%3Aopen+label%3Abug) [](https://gitter.im/Microsoft/vscode) ## The Repository This repository ("`Code - OSS`") is where we (Microsoft) develop the [Visual Studio Code](https://code.visualstudio.com) product together with the community. Not only do we work on code and issues here, but we also publish our [roadmap](https://github.com/microsoft/vscode/wiki/Roadmap), [monthly iteration plans](https://github.com/microsoft/vscode/wiki/Iteration-Plans), and our [endgame plans](https://github.com/microsoft/vscode/wiki/Running-the-Endgame). This source code is available to everyone under the standard [MIT license](https://github.com/microsoft/vscode/blob/main/LICENSE.txt). ## Visual Studio Code <p align="center"> <img alt="VS Code in action" src="https://user-images.githubusercontent.com/35271042/118224532-3842c400-b438-11eb-923d-a5f66fa6785a.png"> </p> [Visual Studio Code](https://code.visualstudio.com) is a distribution of the `Code - OSS` repository with Microsoft-specific customizations released under a traditional [Microsoft product license](https://code.visualstudio.com/License/). [Visual Studio Code](https://code.visualstudio.com) combines the simplicity of a code editor with what developers need for their core edit-build-debug cycle. It provides comprehensive code editing, navigation, and understanding support along with lightweight debugging, a rich extensibility model, and lightweight integration with existing tools. Visual Studio Code is updated monthly with new features and bug fixes. You can download it for Windows, macOS, and Linux on [Visual Studio Code's website](https://code.visualstudio.com/Download). To get the latest releases every day, install the [Insiders build](https://code.visualstudio.com/insiders). ## Contributing There are many ways in which you can participate in this project, for example: * [Submit bugs and feature requests](https://github.com/microsoft/vscode/issues), and help us verify as they are checked in * Review [source code changes](https://github.com/microsoft/vscode/pulls) * Review the [documentation](https://github.com/microsoft/vscode-docs) and make pull requests for anything from typos to and new content. If you are interested in fixing issues and contributing directly to the code base, please see the document [How to Contribute](https://github.com/microsoft/vscode/wiki/How-to-Contribute), which covers the following: * [Ho [ASSISTANT] {"what_it_is": "The open-source core of Visual Studio Code, Microsoft's cross-platform code editor.", "stack": ["TypeScript", "Electron", "Node.js"], "why_mike_cares": "Low overlap as a codebase, but VS Code is the host editor for Claude Code and Codex extensions Mike uses daily.", "verdict": "skip", "verdict_reason": "Mike uses VS Code as a tool, not something he'd fork or extract patterns from at the repo level."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: n8n-io/n8n Stars: 188280 Language: TypeScript Topics: ai, apis, automation, cli, data-flow, development, integration-framework, integrations, ipaas, low-code, low-code-platform, mcp, mcp-client, mcp-server, n8n, no-code, self-hosted, typescript, workflow, workflow-automation Description: Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations. README (first 3000 chars):  # n8n - Secure Workflow Automation for Technical Teams n8n is a workflow automation platform that gives technical teams the flexibility of code with the speed of no-code. With 400+ integrations, native AI capabilities, and a fair-code license, n8n lets you build powerful automations while maintaining full control over your data and deployments.  ## Key Capabilities - **Code When You Need It**: Write JavaScript/Python, add npm packages, or use the visual interface - **AI-Native Platform**: Build AI agent workflows based on LangChain with your own data and models - **Full Control**: Self-host with our fair-code license or use our [cloud offering](https://app.n8n.cloud/login) - **Enterprise-Ready**: Advanced permissions, SSO, and air-gapped deployments - **Active Community**: 400+ integrations and 900+ ready-to-use [templates](https://n8n.io/workflows) ## Quick Start Try n8n instantly with [npx](https://docs.n8n.io/hosting/installation/npm/) (requires [Node.js](https://nodejs.org/en/)): ``` npx n8n ``` Or deploy with [Docker](https://docs.n8n.io/hosting/installation/docker/): ``` docker volume create n8n_data docker run -it --rm --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8n ``` Access the editor at http://localhost:5678 ## Resources - 📚 [Documentation](https://docs.n8n.io) - 🔧 [400+ Integrations](https://n8n.io/integrations) - 💡 [Example Workflows](https://n8n.io/workflows) - 🤖 [AI & LangChain Guide](https://docs.n8n.io/advanced-ai/) - 👥 [Community Forum](https://community.n8n.io) - 📖 [Community Tutorials](https://community.n8n.io/c/tutorials/28) ## Support Need help? Our community forum is the place to get support and connect with other users: [community.n8n.io](https://community.n8n.io) ## License n8n is [fair-code](https://faircode.io) distributed under the [Sustainable Use License](https://github.com/n8n-io/n8n/blob/master/LICENSE.md) and [n8n Enterprise License](https://github.com/n8n-io/n8n/blob/master/LICENSE_EE.md). - **Source Available**: Always visible source code - **Self-Hostable**: Deploy anywhere - **Extensible**: Add your own nodes and functionality [Enterprise Licenses](mailto:license@n8n.io) available for additional features and support. Additional information about the license model can be found in the [docs](https://docs.n8n.io/sustainable-use-license/). ## Cont [ASSISTANT] {"what_it_is":"Fair-code workflow automation platform with 400+ integrations, native AI/LangChain support, MCP client/server nodes, and self-host or cloud deployment.","stack":["TypeScript","Node.js","LangChain","Docker","MCP"],"why_mike_cares":"Direct overlap with Mike's agent ecosystem, MCP work, GoHighLevel/SignalWire/Discord automation, and SEO pipelines, plus n8n is already in his stack mention list.","verdict":"check-it-out","verdict_reason":"n8n is the canonical self-hosted automation hub for AI agent workflows and Mike runs exactly the kind of multi-integration, MCP-driven, self-hosted automation stack n8n is built for."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: ultraworkers/claw-code Stars: 191750 Language: Rust Topics: Description: The repo is finally unlocked. enjoy the party! The fastest repo in history to surpass 100K stars ⭐. Join Discord: https://discord.gg/5TUQKqFWd Built in Rust using oh-my-codex. README (first 3000 chars): # Claw Code <p align="center"> <a href="https://github.com/ultraworkers/claw-code">ultraworkers/claw-code</a> · <a href="./USAGE.md">Usage</a> · <a href="./rust/README.md">Rust workspace</a> · <a href="./PARITY.md">Parity</a> · <a href="./ROADMAP.md">Roadmap</a> · <a href="./CONTRIBUTING.md">Contributing</a> · <a href="./SECURITY.md">Security</a> · <a href="https://discord.gg/5TUQKqFWd">UltraWorkers Discord</a> </p> <p align="center"> <a href="https://star-history.com/#ultraworkers/claw-code&Date"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/svg?repos=ultraworkers/claw-code&type=Date&theme=dark" /> <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/svg?repos=ultraworkers/claw-code&type=Date" /> <img alt="Star history for ultraworkers/claw-code" src="https://api.star-history.com/svg?repos=ultraworkers/claw-code&type=Date" width="600" /> </picture> </a> </p> <p align="center"> <img src="assets/claw-hero.jpeg" alt="Claw Code" width="300" /> </p> Claw Code is the public Rust implementation of the `claw` CLI agent harness. The canonical implementation lives in [`rust/`](./rust), and the current source of truth for this repository is **ultraworkers/claw-code**. > [!IMPORTANT] > Start with [`USAGE.md`](./USAGE.md) for build, auth, CLI, session, and parity-harness workflows. For file submission/navigation questions, see [Navigation and file context](./docs/navigation-file-context.md). For local OpenAI-compatible models and offline skill installs, see [Local OpenAI-compatible providers and skills setup](./docs/local-openai-compatible-providers.md). Windows users can jump to the PowerShell-first [Windows install and release quickstart](./docs/windows-install-release.md). Make `claw doctor` your first health check after building, use [`rust/README.md`](./rust/README.md) for crate-level details, read [`PARITY.md`](./PARITY.md) for the current Rust-port checkpoint, and see [`docs/container.md`](./docs/container.md) for the container-first workflow. > > **ACP / Zed status:** `claw-code` does not ship an ACP/Zed daemon or JSON-RPC entrypoint yet. Run `claw acp` (or `claw --acp`) for the current status instead of guessing from source layout; `claw acp serve` is currently a discoverability alias only, returns status with exit code 0, and real ACP support remains tracked separately in `ROADMAP.md`. For the public JSON contract, see [`docs/g011-acp-json-rpc-status-contract.md`](./docs/g011-acp-json-rpc-status-contract.md). ## Current repository shape - **`rust/`** — canonical Rust workspace and the `claw` CLI binary - **`USAGE.md`** — task-oriented usage guide for the current product surface - **`PARITY.md` [ASSISTANT] {"what_it_is": "Rust-based CLI agent harness (a `claw` CLI) positioned as an alternative coding agent runtime with parity tracking to Codex-style workflows.", "stack": ["Rust", "CLI", "ACP/JSON-RPC", "OpenAI-compatible providers"], "why_mike_cares": "Direct overlap with Mike's agentic coding stack (Claude Code, Codex, OpenClaw) and his interest in CLI-first agent harnesses, though the 191K stars and 'party unlocked' framing smell like a hype/joke repo rather than a serious tool.", "verdict": "worth-knowing", "verdict_reason": "Adjacent to Mike's OpenClaw/Codex/Claude Code ecosystem and CLI-first routing rule, but the suspicious star count and meme-y description suggest validating legitimacy before adopting."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: tensorflow/tensorflow Stars: 195134 Language: C++ Topics: deep-learning, deep-neural-networks, distributed, machine-learning, ml, neural-network, python, tensorflow Description: An Open Source Machine Learning Framework for Everyone README (first 3000 chars): <div align="center"> <img src="https://www.tensorflow.org/images/tf_logo_horizontal.png"> </div> [](https://badge.fury.io/py/tensorflow) [](https://badge.fury.io/py/tensorflow) [](https://doi.org/10.5281/zenodo.4724125) [](https://bestpractices.coreinfrastructure.org/projects/1486) [](https://securityscorecards.dev/viewer/?uri=github.com/tensorflow/tensorflow) [](https://bugs.chromium.org/p/oss-fuzz/issues/list?sort=-opened&can=1&q=proj:tensorflow) [](https://bugs.chromium.org/p/oss-fuzz/issues/list?sort=-opened&can=1&q=proj:tensorflow-py) [](https://ossrank.com/p/44) [](CODE_OF_CONDUCT.md) **`Documentation`** | ------------------- | [](https://www.tensorflow.org/api_docs/) | [TensorFlow](https://www.tensorflow.org/) is an end-to-end open source platform for machine learning. It has a comprehensive, flexible ecosystem of [tools](https://www.tensorflow.org/resources/tools), [libraries](https://www.tensorflow.org/resources/libraries-extensions), and [community](https://www.tensorflow.org/community) resources that lets researchers push the state-of-the-art in ML and developers easily build and deploy ML-powered applications. TensorFlow was originally developed by researchers and engineers working within the Machine Intelligence team at Google Brain to conduct research in machine learning and neural networks. However, the framework is versatile enough to be used in other areas as well. TensorFlow provides stable [Python](https://www.tensorflow.org/api_docs/python) and [C++](https://www.tensorflow.org/api_docs/cc) APIs, as well as a non-guaranteed backward compatible API for [other languages](https://www.tensorflow.org/api_docs). Keep up-to-date with release announcements and security updates by subscribing to [announce@tensorflow.org](https://groups.google.com/a/tensorflow.org/forum/#!forum/announce). See all the [mailing lists](https://www.tensorflow.org/community/forums). ## Install See the [TensorFlow install guide](https://www.tensorflow.org/install) for the [pi [ASSISTANT] { "what_it_is": "Google's end-to-end open source machine learning framework for building and deploying neural networks.", "stack": ["Python", "C++", "CUDA", "Keras"], "why_mike_cares": "Low overlap - Mike consumes LLM APIs (Claude, Gemini) rather than training models from scratch.", "verdict": "skip", "verdict_reason": "Foundational ML framework, not agentic tooling - Mike's stack is API-driven LLM orchestration, not model training infrastructure." }

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: vinta/awesome-python Stars: 298097 Language: Python Topics: awesome, collections, python, python-frameworks, python-libraries, python-tools Description: An opinionated list of Python frameworks, libraries, tools, and resources README (first 3000 chars): # [Awesome Python](https://awesome-python.com/) An opinionated guide to the best Python frameworks, libraries, tools, and resources. **Visit the [website](https://awesome-python.com/) to search and filter projects more easily.** ## **Sponsors** - **[pyr](https://pyrun.dev)**: Zero-config Python project manager. Bootstraps its own runtime, app-convention, and working imports - out the box. > The **#10 most-starred repo on GitHub**. Put your product in front of Python developers. [Become a sponsor](SPONSORSHIP.md). ## Categories **AI & ML** - [AI and Agents](#ai-and-agents) - [Deep Learning](#deep-learning) - [Machine Learning](#machine-learning) - [Natural Language Processing](#natural-language-processing) - [Computer Vision](#computer-vision) - [Recommender Systems](#recommender-systems) **Web Development** - [Web Frameworks](#web-frameworks) - [Web APIs](#web-apis) - [Web Servers](#web-servers) - [WebSocket](#websocket) - [Template Engines](#template-engines) - [Web Asset Management](#web-asset-management) - [Authentication](#authentication) - [Admin Panels](#admin-panels) - [CMS](#cms) - [Static Site Generators](#static-site-generators) **HTTP & Scraping** - [HTTP Clients](#http-clients) - [Web Scraping](#web-scraping) - [Email](#email) **Database & Storage** - [ORM](#orm) - [Database Drivers](#database-drivers) - [Database](#database) - [Caching](#caching) - [Search](#search) - [Serialization](#serialization) **Data & Science** - [Data Analysis](#data-analysis) - [Data Ingestion / ETL](#data-ingestion--etl) - [Data Validation](#data-validation) - [Data Visualization](#data-visualization) - [Geolocation](#geolocation) - [Science](#science) - [Quantum Computing](#quantum-computing) **Developer Tools** - [Algorithms and Design Patterns](#algorithms-and-design-patterns) - [Interactive Interpreter](#interactive-interpreter) - [Code Analysis](#code-analysis) - [Testing](#testing) - [Debugging Tools](#debugging-tools) - [Build Tools](#build-tools) - [Documentation](#documentation) **DevOps** - [DevOps Tools](#devops-tools) - [Distributed Computing](#distributed-computing) - [Task Queues](#task-queues) - [Messaging](#messaging) - [Job Schedulers](#job-schedulers) - [Logging](#logging) - [Network Virtualization](#network-virtualization) **CLI & GUI** - [CLI Development](#cli-development) - [CLI Tools](#cli-tools) - [GUI Development](#gui-development) **Text & Documents** - [Text Processing](#text-processing) - [HTML Manipulation](#html-manipulation) - [File Format Processing](#file-format-processing) - [File Manipulation](#file-manipulation) **Media** - [Image Processing](#image-processing) - [Audio & Video Processing](#audio--video-processing) - [Game Development](#game-development) **Python Language** - [Implementations](#implementat [ASSISTANT] {"what_it_is": "A curated mega-list of Python frameworks, libraries, and tools organized by category.", "stack": ["Python", "Markdown"], "why_mike_cares": "Useful reference catalog for picking Python libs across AI/ML, web scraping, task queues, and CLI dev that Mike uses in automation pipelines.", "verdict": "worth-knowing", "verdict_reason": "Not actionable on its own, but a solid lookup index when scoping Python tooling for agents, scrapers, or schedulers."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: openclaw/openclaw Stars: 372550 Language: TypeScript Topics: ai, assistant, crustacean, molty, openclaw, own-your-data, personal Description: Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞 README (first 3000 chars): # 🦞 OpenClaw — Personal AI Assistant <p align="center"> <picture> <source media="(prefers-color-scheme: light)" srcset="https://raw.githubusercontent.com/openclaw/openclaw/main/docs/assets/openclaw-logo-text-dark.svg"> <img src="https://raw.githubusercontent.com/openclaw/openclaw/main/docs/assets/openclaw-logo-text.svg" alt="OpenClaw" width="500"> </picture> </p> <p align="center"> <strong>EXFOLIATE! EXFOLIATE!</strong> </p> <p align="center"> <a href="https://github.com/openclaw/openclaw/actions/workflows/ci.yml?branch=main"><img src="https://img.shields.io/github/actions/workflow/status/openclaw/openclaw/ci.yml?branch=main&style=for-the-badge" alt="CI status"></a> <a href="https://github.com/openclaw/openclaw/releases"><img src="https://img.shields.io/github/v/release/openclaw/openclaw?include_prereleases&style=for-the-badge" alt="GitHub release"></a> <a href="https://discord.gg/clawd"><img src="https://img.shields.io/discord/1456350064065904867?label=Discord&logo=discord&logoColor=white&color=5865F2&style=for-the-badge" alt="Discord"></a> <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-blue.svg?style=for-the-badge" alt="MIT License"></a> </p> **OpenClaw** is a _personal AI assistant_ you run on your own devices. It answers you on the channels you already use. It can speak and listen on macOS/iOS/Android, and can render a live Canvas you control. The Gateway is just the control plane — the product is the assistant. If you want a personal, single-user assistant that feels local, fast, and always-on, this is it. Supported channels include: WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, IRC, Microsoft Teams, Matrix, Feishu, LINE, Mattermost, Nextcloud Talk, Nostr, Synology Chat, Tlon, Twitch, Zalo, Zalo Personal, WeChat, QQ, WebChat. [Website](https://openclaw.ai) · [Docs](https://docs.openclaw.ai) · [Vision](VISION.md) · [DeepWiki](https://deepwiki.com/openclaw/openclaw) · [Getting Started](https://docs.openclaw.ai/start/getting-started) · [Updating](https://docs.openclaw.ai/install/updating) · [Showcase](https://docs.openclaw.ai/start/showcase) · [FAQ](https://docs.openclaw.ai/help/faq) · [Onboarding](https://docs.openclaw.ai/start/wizard) · [Nix](https://github.com/openclaw/nix-openclaw) · [Docker](https://docs.openclaw.ai/install/docker) · [Discord](https://discord.gg/clawd) New install? Start here: [Getting started](https://docs.openclaw.ai/start/getting-started) Preferred setup: run `openclaw onboard` in your terminal. OpenClaw Onboard guides you step by step through setting up the gateway, workspace, channels, and skills. It is the recommended CLI setup path and works on **macOS, Linux, and Windows (via WSL2; strongly recommended)**. Works with npm, pnpm, or bun. ## Sponsors <t [ASSISTANT] {"what_it_is":"OpenClaw is a self-hosted personal AI assistant that runs across 20+ chat channels (WhatsApp, Telegram, Slack, Discord, iMessage, etc.) with voice, canvas, and skills/plugins.","stack":["TypeScript","Node.js","Discord API","Telegram API","Slack API","WhatsApp","Docker","Nix"],"why_mike_cares":"Mike already operates OpenClaw in his ecosystem (OliverOscar agent, Mac-side orchestration, Telegram/Discord routing, SecureClaw, skill/plugin vetting) and has a dedicated openclaw-vps-operator skill plus Matteo-Jr watching its runtime.","verdict":"check-it-out","verdict_reason":"This is a core production runtime in Mike's stack, not a curiosity — any upstream change directly impacts his agent fleet."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: freeCodeCamp/freeCodeCamp Stars: 444977 Language: TypeScript Topics: careers, certification, community, curriculum, d3, education, freecodecamp, javascript, learn-to-code, math, nodejs, nonprofits, programming, react, teachers Description: freeCodeCamp.org's open-source codebase and curriculum. Learn math, programming, and computer science for free. README (first 3000 chars): [](https://www.freecodecamp.org/) [](https://www.firsttimersonly.com/) [](https://discord.gg/PRyKn3Vbay) [](https://insights.linuxfoundation.org/project/freecodecamp/repository/freecodecamp-freecodecamp) ## freeCodeCamp.org's open-source codebase and curriculum [freeCodeCamp.org](https://www.freecodecamp.org) is a friendly community where you can learn to code for free. It is run by a [donor-supported 501(c)(3) charity](https://www.freecodecamp.org/donate) to help millions of busy adults transition into tech. Our community has already helped more than 100,000 people get their first developer job. Our full-stack web development and machine learning curriculum is completely free and self-paced. We have thousands of interactive coding challenges to help you expand your skills. ## Table of Contents - [Certifications](#certifications) - [The Learning Platform](#the-learning-platform) - [Reporting Bugs and Issues](#reporting-bugs-and-issues) - [Reporting Security Issues and Responsible Disclosure](#reporting-security-issues-and-responsible-disclosure) - [Contributing](#contributing) - [Platform, Build and Deployment Status](#platform-build-and-deployment-status) - [License](#license) ### Certifications freeCodeCamp.org offers several free developer certifications that make up the [Full-Stack Developer Curriculum](https://www.freecodecamp.org/learn/full-stack-developer-v9/): - [Responsive Web Design](https://www.freecodecamp.org/learn/responsive-web-design-v9/) - [JavaScript](https://www.freecodecamp.org/learn/javascript-v9/) - [Front-End Development Libraries](https://www.freecodecamp.org/learn/front-end-development-libraries-v9/) - [Python](https://www.freecodecamp.org/learn/python-v9/) - [Relational Databases](https://www.freecodecamp.org/learn/relational-databases-v9/) - [Back-End Development and APIs](https://www.freecodecamp.org/learn/back-end-development-and-apis-v9/) Each of these certifications involves completing interactive lessons, workshops, labs, reviews, and quizzes. Throughout the certification, you'll need to complete 5 required projects to qualify for the exam. Once you pass the exam, then you can claim the certification. freeCodeCamp.org also offers free langu [ASSISTANT] {"what_it_is": "freeCodeCamp's open-source learning platform and curriculum for teaching coding, math, and computer science to millions of self-taught developers.","stack": ["TypeScript","JavaScript","React","Node.js","D3"],"why_mike_cares": "Low overlap, this is a learn-to-code education platform, not agent tooling or SEO automation.","verdict": "skip","verdict_reason": "Pure learning curriculum with no connection to Mike's AI agent, SEO automation, or agency tooling work."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: opensquilla/opensquilla Stars: 937 Language: Python Topics: agent, ai, ai-agents, deep-learning, foundation-models, llm, mcp, memory, openclaw, python, skills Description: OpenSquilla — Token-Efficient AI Agent with same budget, higher intelligence density README (first 3000 chars): # OpenSquilla — Token-Efficient AI Agent <p align="center"> <img src="assets/opensquilla-long-logo.png" alt="OpenSquilla logo" width="500"> </p> <p align="center"> <a href="https://opensquilla.ai/"><img src="https://img.shields.io/badge/website-opensquilla.ai-blue?style=for-the-badge" alt="Website"></a> <a href="https://github.com/opensquilla/opensquilla/releases"><img src="https://img.shields.io/github/v/release/opensquilla/opensquilla?include_prereleases&style=for-the-badge" alt="GitHub release"></a> <a href="LICENSE"><img src="https://img.shields.io/badge/License-Apache%202.0-blue.svg?style=for-the-badge" alt="Apache 2.0 License"></a> </p> ## Overview OpenSquilla is a token-efficient, microkernel AI agent — same budget, more capability, better results. It combines smart routing, persistent memory, a secure sandbox, built-in web search, and local embeddings under a single model loop. Every entry point — Web UI, CLI, and chat channels — runs through a shared `TurnRunner`, and a pluggable provider layer lets it speak to OpenRouter, OpenAI, Anthropic, Ollama, DeepSeek, Gemini, Qwen/DashScope, and roughly twenty other LLM providers without changes to your code or config schema. ## Quick start Choose the path that matches how you want to use OpenSquilla: | User type | Path | Status | | --- | --- | --- | | New user | [Preview release package](#preview-release-package) | Recommended | | Command-line user | [Install from source](#install-from-source) | Available now | | Developer | [Develop from source](#develop-from-source) | Available now | SquillaRouter is included by default in the preview release packages and in the normal source install path. Only choose the `core` profile or `--router disabled` if you intentionally want to skip the bundled router. ### Preview release package Download the preview package if you want to try OpenSquilla as a local app without cloning the repository or installing Git, Git LFS, or `uv`. 1. Download the package from the [GitHub Releases](https://github.com/opensquilla/opensquilla/releases) page and extract it to a writable folder. 2. Double-click `Start OpenSquilla.cmd` from the extracted folder. Keep the terminal window open. Closing it stops the gateway. 3. Complete onboarding and open the Web UI. The launcher opens onboarding before the gateway starts. On first run, choose a provider and paste the requested keys; later starts let you review or change the config. Then open <http://127.0.0.1:18790/control/>. <details> <summary>Advanced portable usage</summary> Use these options only when you want scripted setup or portable CLI commands. - To provide an OpenRouter key before first start: ```powershell $env:OPENROUTER_API_KEY="sk-..." Set-ExecutionPo [ASSISTANT] { "what_it_is": "Token-efficient microkernel AI agent with smart routing, persistent memory, sandbox, and 20+ LLM provider support under a unified TurnRunner.", "stack": ["Python", "OpenRouter", "OpenAI", "Anthropic", "Ollama", "MCP", "FastAPI"], "why_mike_cares": "Direct overlap with OpenClaw ecosystem (topic tag confirms it), agent skills, MCP, memory layers, and multi-provider routing Mike already runs across Claude Code, Codex, Gemini, and DeepSeek.", "verdict": "check-it-out", "verdict_reason": "OpenClaw-tagged agent framework with skills, MCP, memory, and provider abstraction maps 1:1 onto Mike's active OpenClaw VPS/Mac fleet and Hindsight memory work." }

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: HermannBjorgvin/Clawdmeter Stars: 1108 Language: C Topics: Description: ESP32 desk dashboard that shows Claude Code usage README (first 3000 chars): Analyze this repo per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] {"what_it_is":"ESP32-based physical desk dashboard that displays real-time Claude Code usage stats on a small hardware display.","stack":["ESP32","C","Claude Code","Arduino"],"why_mike_cares":"Mike runs Claude Code as his primary agent across multiple machines and tracks usage/token spend obsessively, a physical desk monitor for Claude usage fits his ClawControl/observability obsession.","verdict":"worth-knowing","verdict_reason":"Novel hardware angle on Claude Code observability that complements his existing dashboards, but it's a hobby ESP32 build not core to his Next.js/agent stack."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: yetone/native-feel-skill Stars: 1254 Language: Topics: Description: An Agent Skill for designing cross-platform desktop apps that feel native — distilled from Raycast's 2.0 deep-dive and reverse engineering of Raycast Beta.app. Eight architectural tenets, four-layer architecture, WebKit/WebView2 survival guide, 75-item ship audit. README (first 3000 chars): Analyze this repo per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] {"what_it_is":"An Agent Skill that teaches Claude/agents how to design cross-platform desktop apps that feel truly native, distilled from Raycast 2.0 reverse engineering with 8 architectural tenets, a 4-layer architecture, and a 75-item ship audit.","stack":["Agent Skills","Claude Code","WebKit","WebView2","Cross-platform desktop"],"why_mike_cares":"Mike lives in the Agent Skills ecosystem and routinely ships Electron/WebView-style dashboards (ClawControl, HawkeyePanel, Hermes MCC), so a battle-tested skill for native-feel desktop UX plugs directly into his stack.","verdict":"check-it-out","verdict_reason":"High-signal Agent Skill with 1.2k stars that upgrades any of Mike's desktop/dashboard builds and slots straight into his Claude Code skill library."}

[USER] You are an expert technical analyst for Mike Merlino, an AI agency operator and builder. Your job is to evaluate GitHub repositories and extract structured signal. Mike's domains: AI agents, LLM tooling, Claude Code, MCP servers, prompt engineering, SEO automation, GoHighLevel, Next.js, ShadCN, Vercel, Supabase, voice AI, agentic coding, Discord/Telegram bots, scheduler/cron patterns, Python automation, TypeScript. Return ONLY valid JSON: { "what_it_is": "one sentence plain English description", "stack": ["Technology1", "Technology2"], "why_mike_cares": "one sentence on overlap with Mike's work, or 'Low overlap'", "verdict": "check-it-out" | "worth-knowing" | "skip", "verdict_reason": "one sentence" } Verdict: - "check-it-out": direct overlap with Mike's active projects or tools he uses - "worth-knowing": interesting adjacent tool, may matter later - "skip": no clear overlap Be OPINIONATED. Not neutral. Repo: vercel-labs/zero Stars: 1300 Language: C Topics: Description: The programming language for agents README (first 3000 chars): Analyze this repo per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] { "what_it_is": "Vercel Labs' experimental programming language designed specifically for building AI agents.", "stack": ["C", "DSL", "Agent Runtime"], "why_mike_cares": "Mike builds agentic systems daily (Claude Code, Oliver, Carlos, OpenClaw) and a Vercel-backed agent DSL could shape how agent orchestration is written.", "verdict": "worth-knowing", "verdict_reason": "Early-stage agent language from Vercel Labs with no clear production fit yet, but worth tracking since Mike's stack is Vercel-heavy and agent-first." }

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: The Next New Thing AI Video title: TradingAgents: AI Agents Are Changing Finance TRANSCRIPT (first 6000 chars): Trading agents. This is a tool that helps people trade. This guy used it and he showed how it analyzed Nvidia. It told him that from lots of different directions, lots of different user profiles, and it ended up with a target of 268. And then he looked up how much Nvidia was at the time that he did it, and it was 199. This is basically a trend that we're seeing, right? AI is coming for finance. People are thinking that they could use it to make better investment decisions. Smart, useful. What? >> Yeah, I think I think it's interesting. You know, AI is very good at creating models. We do that at our studio all the time for our businesses. >> The big value I see in this is you've got research that you can use, whether or not you decide to to invest in this. Download it in the link in the bio. Extract the structured JSON signal per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] {"tldr":["TradingAgents is a multi-agent AI system that analyzes stocks from multiple perspectives/user profiles to generate price targets (demo showed NVDA target of $268 when price was $199)","Pitched as AI-driven finance research tool, not a trading executor — outputs reasoning and targets you can act on or ignore","Channel framing is promotional/affiliate-style (download in bio), so treat the NVDA hit as anecdote not validation","No technical depth on the agent architecture, prompts, or how the user profiles are constructed","Outside Mike's domain stack — no overlap with agency, SEO, local, or agentic coding workflows"],"tools":[{"name":"TradingAgents","url":"https://github.com/TauricResearch/TradingAgents","description":"Multi-agent LLM framework that simulates a trading firm (analysts, researchers, traders) to produce stock analysis and price targets"}],"skill_candidates":[],"verdict":"skip","verdict_reason":"Finance/trading agent demo with no Mike-domain overlap, no novel agent pattern disclosed, and promotional framing — zero actionable takeaway for the agency stack."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: The Next New Thing AI Video title: Look up anyone across 3,000 sites — for FREE | Maigret TRANSCRIPT (first 6000 chars): Dude, this is going to look anyone up across 3,000 sites. Facebook, Instagram, GitHub. Just basically if you want to stalk somebody, that's what it's for. What's a What's a good use of this? >> a That's a negative spin on it, you know? The a lot of people in your audience and our like studio businesses. Look, we're trying to find leads. We're trying to find sales leads. We're trying to find talent. We're trying to understand people and I would much rather deeply understand a person before I reach out to them to sell them something or to hire them. So, we do this already, but honestly, we do compared to this what looks like a terrible tiny version of it. We're scraping, you know, LinkedIn, Twitter, Reddit. So, I'm very curious to try this out to basically enrich our lead at a higher level. I'm curious how expensive it will be. >> It looks like a lot of AI's going to have to run to scrape all these things, but I Absolutely, I will try this this week. I get it. So, you're saying, "Look, we're thinking of investing in someone or hiring someone. I'm already looking them up on LinkedIn. That's the baby version of this. I want the the bigger thing. What are they doing on GitHub? What did they post on Reddit? Let's really understand the person not for nefarious reasons." Okay, fair. Great great tool here. Download it in the link in the bio. Extract the structured JSON signal per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] {"tldr":["Maigret is an OSINT tool that scans 3,000+ sites (Facebook, Instagram, GitHub, Reddit, Twitter) for a single username/identity","Use case for Mike: lead enrichment beyond LinkedIn scraping — pull GitHub, Reddit, forum activity to deeply qualify prospects before outreach","Positioned as upgrade over manual LinkedIn/Twitter/Reddit scraping currently used for sales lead enrichment","Free and open-source (Python CLI) — runs locally, no per-lookup API cost","Pairs well with cold outreach workflows: enrich lead → personalize SMS/email → contact"],"tools":[{"name":"Maigret","url":"https://github.com/soxoj/maigret","description":"Free OSINT Python CLI that collects a person's dossier from 3,000+ sites by username"}],"skill_candidates":[{"slug":"lead-enrichment-maigret","description":"Take a name/username/email, run Maigret across 3,000+ sites, aggregate GitHub/Reddit/forum signals into a lead dossier for cold outreach personalization"}],"verdict":"dont-miss","verdict_reason":"Free OSINT tool that hits Mike's cold outreach + Python automation domains and slots directly into existing lead enrichment workflows feeding SMS/GoHighLevel campaigns."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Julian Goldie SEO Video title: New 0GM-1.0-35B-A3B AI Agent is INSANE! 🤯 TRANSCRIPT (first 6000 chars): ZojiM 1.035 BA3B's insane. And most people have no idea this just dropped. This is one of those releases that looks quiet on the surface. No massive PR campaign, no big influencer push, but when you actually read what's under the hood, it hits different. So, let's get into it. Zoji Labs just released their first ever proprietary AI model. It's called ZojiM 1.035B A3B. And the thing that makes it wild is not just the model itself. It's the whole story behind how it was built, what it was built for, and what it means for where AI is going. Here's the quick version. Parameters. Around 3 billion active at any given time. Context window of 262,000 tokens, extendable up to 1 million. Built for agents, tool use, and long reasoning chains, and open sourced under Apache 2.0, meaning you can use it, fine-tune it, build products with it, and run it yourself. No restrictions. Hey, if we haven't met already, I'm the digital avatar of Julian Goldie, CEO of SEO agency Goldie Agency. Whilst he's helping clients get more leads and customers, I'm here to help you get the latest AI updates. And Julian Goldie reads every comment, so make sure you comment below. Here's the thing nobody's really talking about. It was trained on a decentralized GPU network. Not AWS, not Google Cloud, not some massive centralized AI data center in Silicon Valley. The actual training of this model happened across decentralized GPUs, Zoji's own compute network, and then it was deployed on their own infrastructure. And now it runs in their private computer platform. The whole loop is closed. Trained here, served here, and here. That's genuinely new. Most AI models in the Web3 space are just regular models from OpenAI or one of the big labs wrapped in token economics. The chain handles payment. The actual intelligence still lives inside one company's server. It's different. The intelligence layer itself runs on decentralized infrastructure. Now, let me explain the technical side simply, because the architecture is what makes this actually useful. Uses something called mixture of experts. For short. Here's the simple version. Imagine a company with 256 specialists. Every time a task comes in, instead of waking up all 256 people, the system picks the eight or nine most relevan [ASSISTANT] {"tldr":["Zoji Labs dropped ZojiM 1.0-35B-A3B, an Apache 2.0 open-source MoE model (35B total / 3B active) built specifically for agentic workflows","Native 262K context window extendable to 1M tokens, making it viable for holding entire business knowledge bases (SOPs, client notes, email history) in one session","Reasoning is on by default via 'think tags' before final output, plus native tool-use chaining for multi-step agent tasks","Trained AND served on Zoji's own decentralized GPU network, unlike Web3 AI wrappers that still rely on OpenAI underneath","Rest of video is a pitch for Julian's AI Profit Boardroom paid community, no concrete implementation details"],"tools":[{"name":"ZojiM 1.0-35B-A3B","url":"https://zoji.ai","description":"Open-source MoE agentic model with 262K-1M context, built for tool use and long reasoning chains"}],"skill_candidates":[],"verdict":"skip","verdict_reason":"Surface-level model announcement with no benchmarks, no install/usage demo, no code, and most of the runtime is a paid community pitch — Mike can grab the model name and check HuggingFace directly in 30 seconds."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Julian Goldie SEO Video title: Grok Build AI is WILD: Automate ANYTHING! TRANSCRIPT (first 6000 chars): Grok Build AI is wild, automate anything. Grok Build AI is wild, and by the end of this you're going to see exactly why XAI just flipped the script on how people build things with AI. So, here's what this actually is. Grok Build is XAI's new agentic CLI tool. Type commands into a terminal, that black window on your computer, and what Grok Build does inside that terminal is insane. Doesn't just write code, plans, builds, runs, fixes its own mistakes, loops until the job is done. You give it a goal and it figures out how to get there. That's different from what we've had before. State AI coding tools sit next to you, suggest things, you still do the work. Grok Build actually does the work. You're more like the person giving the brief, and Grok Build is the one actually building. And here's the part that makes this relevant to you even if you've never touched a terminal in your life. This thing can automate workflows, real ones, not toy examples, full sequences of tasks that used to take hours, or that you'd have to hire someone to do. Let me break down how it actually works because this is where it gets interesting. Grok Build runs on Grok 3. That's XAI's most powerful model right now. And the way it's built, it runs in what's called an agentic loop. That means it doesn't just answer once and stop. It acts, checks the result, decides what to do next, acts again. It keeps going until it hits the goal you gave it, or until it hits a wall and asks you what you want to do. It has access to your file system. It can read your files, write new ones, edit existing ones, run terminal commands on your machine, search the web, external APIs, and it does all of this with what XAI calls a minimal footprint. It asks for permission before doing anything big. Terrified it's going to delete everything, it checks in. That's a really important design choice because a lot of people are nervous about giving AI this kind of access. Guardrails that keep you in the loop without slowing the whole thing down. Here's a real example of what this looks like in practice. A community, like let's use the AI Profit Boardroom as the example because it's a real business that does real things. You want to build an automated onboarding sequence. Member joins, gets a we [ASSISTANT] { "tldr": [ "xAI shipped Grok Build — an agentic CLI that plans, writes, runs, and self-corrects code in a loop until the goal is hit, with permission gates before destructive actions.", "Runs on Grok 3 with filesystem access, terminal exec, web search, and external API calls — positioned as a direct competitor to Claude Code and Codex CLI.", "Available via xAI API, so it can be piped into existing automation stacks instead of living only in a terminal window.", "Use cases pitched: community onboarding workflows, content repurposing pipelines (blog → Twitter/LinkedIn/email), automated client reporting, CRM inactive-lead flagging.", "Heavy AI Profit Boardroom upsell — the playbook content itself is thin; the real signal is 'another agentic CLI exists, evaluate vs Claude Code.'" ], "tools": [ { "name": "Grok Build", "url": "https://x.ai", "description": "xAI's agentic CLI tool running on Grok 3 — plans, codes, executes, self-corrects in a loop with filesystem and terminal access." } ], "skill_candidates": [ { "slug": "agentic-cli-comparison", "description": "Side-by-side eval framework for agentic CLIs (Claude Code, Codex, Gemini CLI, Grok Build) — same task, same repo, scored on completion rate, reliability, cost, permission model." } ], "verdict": "worth-a-skim", "verdict_reason": "New agentic CLI in Mike's exact lane (Claude Code competitor) is worth knowing exists, but the video is a thin product announcement wrapped in a community upsell with no novel patterns or hands-on depth." }

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Julian Goldie SEO Video title: Claude Opus 4.7 + NotebookLM is INSANE! TRANSCRIPT (first 6000 chars): Claude Opus 4.7 plus notebook LM is insane. Today I'm going to show you a brand new SEO workflow that's blowing up right now. I'll show you how to use Claude Opus 4.7 with notebook LM to build a full SEO system in three prompts. Most people are doing SEO the hard way. Won't. Let's go. So, here's the deal. Most people think AI is just a chatbot. Type a question. Get an answer. That's it. But the real power isn't in one tool. It's in chaining two tools together. And that's exactly what we're doing today. Claude Opus 4.7 is Anthropic's newest model. It's better at long tasks. It follows instructions more carefully. And it even checks its own work before it answers you. That last part is huge. Because when you're doing SEO work, you don't want the AI guessing. You want the AI thinking. Notebook LM is Google's research tool. You upload your stuff into it. PDFs, posts, articles, whatever. And it reads everything. Then it answers based on what you gave it. Doesn't make stuff up. It pulls from your sources. That's why SEO people are obsessed with it right now. Hey, if we haven't met already, I'm the digital avatar of Julian Goldie, CEO of SEO agency Goldie Agency. Whilst he's helping clients get more leads and customers, I'm here to help you get the latest AI updates. Julian Goldie reads every comment, so make sure you comment below. Here's where it gets fun. When you connect these two tools, you basically build a mini SEO team. One AI does the deep thinking. The other AI organizes everything. Together, they can build out a full content plan in minutes, not days. Minutes. Let me show you how this works for a real business. I'll use my own community as the example, the AI Profit Boardroom. That way you can copy exactly what I'm doing for your business, too. Step one. Open Claude Opus 4.7. Going to do the research first. This is the brain work. Here's the prompt I use. Research the best SEO strategy for an AI automation community called the AI Profit Boardroom. Look at what people are searching for. Find keyword gaps. Find content angles. And what competitors are missing. And give me a list of 20 blog post ideas that would rank fast and bring in real members. I hit enter and watch what happens. Claude Opus 4.7 will start thinking. It plans th [ASSISTANT] {"tldr":["Chain Claude Opus 4.7 (deep research) into NotebookLM (source-grounded filtering) then back into Claude for writing — 3-prompt loop produces a full blog post in ~10 min","Workflow extends to landing pages, keyword clusters, and topic maps by reusing the same Claude→NotebookLM→Claude pattern","NotebookLM's source-only constraint kills hallucination on the filtering step — that's the actual unlock, not the model upgrade","Julian's example prompts are usable as-is: research → top-5 picks with intent+angle+outline → finished post prompt","Mostly promotional for his SEO agency; the technique itself is the takeaway, not the demo"],"tools":[{"name":"Claude Opus 4.7","url":"https://claude.ai","description":"Anthropic's reasoning model used for SEO research and long-form writing"},{"name":"NotebookLM","url":"https://notebooklm.google.com","description":"Google's source-grounded research tool that filters/organizes uploaded material without hallucinating"}],"skill_candidates":[{"slug":"claude-notebooklm-seo-loop","description":"3-prompt chain: Claude researches SEO strategy → NotebookLM filters to top picks with intent+angle+outline → Claude writes the finished asset (blog, landing page, cluster map)"}],"verdict":"worth-a-skim","verdict_reason":"Hits SEO automation + Claude domains and the source-grounded-filter pattern is a clean reusable loop, but no new tools and the technique is a basic two-tool chain Mike already runs more sophisticated versions of."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Julian Goldie SEO Video title: Hermes Agent Just Did Something That Scared Me TRANSCRIPT (first 6000 chars): Hermes agent just got a massive upgrade and when you combine it with Gro OS, your AI agent becomes one of the most powerful setups on the planet right now. We're talking about real time Twitter search, image generation, video generation, voice, all inside one system. And if you're already on X, this is actually free with your subscription already. And here's the thing most people don't realize yet. This isn't just another update. This is your AI agent finally getting eyes, ears, and a voice all at the same time. Now, in this video, I'm going to show you how to step by step plug Grock directly into Hermes agent so it can search the internet in real time. It can create images and videos on demand and talk back to you out loud. And I'm going to show you the one thing you absolutely have to set up after you connect these two together because without it, you're leaving most of the power on the table. So stick with me to the end because by the time we're done, your Hermes agent is going to be running on a completely different level. Let's get into it. So today we're going to be talking about how to use the new update from Grock inside Hermes agent. This creates and unlocks some really powerful features. And basically what's happened here is you can now use X directly with your subscription inside Hermes agent. Right? So this means for example you can use this for searching X, you can use it to generate images, you can use it for video generation and you can also use Hermes agent with Grock for voice too. So this unlocks a whole new bunch of powerful features that I'm going to cover with you today and show you exactly how to use it directly. It's a brand new update that makes Hermes agent way more powerful than ever before. So what you can see here is I've actually plugged this into my agent OS link in the comments in description if you want to get this and you can see that we are using gro 4.3 with grock oorth inside Hermes agent right super easy to add now anytime I update the model it will plug straight into the operating system that we have right here so now I can use Hermes agent to for example generate images or generate videos. So let's test this out. I'm going to say, okay, search X for trending AI automation news. We'll plu [ASSISTANT] {"tldr":["Julian shows how to swap the model inside 'Hermes agent' (his branded CLI wrapper) to xAI Grok 4.3 via `hermes model` so X subscribers get Grok search/image/video/voice with no extra cost","Workflow: run `hermes update` first, then `hermes model` → select X AI Grok, log in via browser popup, then enable `X search` inside `hermes tools` (spacebar to toggle)","Hermes didn't natively know how to call Grok's X search until he pasted Grok/X docs into the agent and trained it on the endpoint — classic 'feed the docs' pattern","Pitched as plug-and-play real-time Twitter/X search + multimodal gen inside one CLI agent, gated behind an X Premium sub","Heavy affiliate-flavored demo (Agent OS link in description); actual novel technical content is thin"],"tools":[],"skill_candidates":[],"verdict":"skip","verdict_reason":"Affiliate-driven demo of a branded CLI wrapper with no verifiable URLs, no new tool Mike doesn't have, and the only real pattern (paste docs to teach an agent a new API) is already standard practice in his stack."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Julian Goldie SEO Video title: NEW Googlebook AI changes everything! TRANSCRIPT (first 6000 chars): New Google Book, AI changes everything, and I mean that literally. Google just built a laptop where Gemini isn't a tab you open. It's not a button you press. It's running in the background, watching your screen, understanding your context, and ready to help before you even ask. This is the Google Book, and it's a completely different kind of machine. Let me break down what's actually going on here, what this thing can do, and why it matters for how you work every single day. Here's the thing about every AI laptop that came before this. They were normal laptops with an AI assistant bolted on. You'd open a chat box, type a question, get an answer, close it, go back to work. The AI had no idea what was on your screen, had no idea what you were doing. It was completely disconnected from your actual workflow. Google Book breaks that pattern completely. Gemini on the Google Book has what Google calls always-on context awareness. That means it sees your screen, understands what you're working on, and it can help you with the exact thing you're doing right now without you explaining anything. You're writing an email and you need to find something you said in a meeting last week. Gemini finds it. You're on a call and someone mentions a company. Gemini pulls up the info. You're working in a Google Doc and you want to rewrite a section. Gemini does it inside the Doc. No copy-paste, switching tabs, no explaining yourself. This is what AI-native actually means. The AI is built into the operating layer of the machine. It's not an app, it's the environment. Now, here's where it gets really interesting. Gemini on the Google Book isn't just reactive, it's proactive. Google has built what they're calling an ambient intelligence layer. So, as you work, Gemini is quietly learning your patterns, what you work on, when, how you like things structured, and it starts surfacing suggestions before you even know you need them. Think about what that means in a real work context. You open your laptop Monday morning. Gemini already knows you have a client meeting at 10. It's pulled the relevant emails, the last meeting notes, and the shared doc you were both working on. Summarized the key points. It's drafted three agenda items based on what's been discussed befor [ASSISTANT] {"tldr":["Google launched 'Google Book' laptop with Gemini running ambient in the OS, not as a bolted-on chat app","Gemini has always-on screen context awareness plus persistent memory across sessions, learning workflow patterns over time","Runs Gemini Ultra on-device for fast/private tasks, seamlessly handing off heavier work to cloud","Native integration across Gmail, Calendar, Drive, Meet, Docs gives Google an integration moat Apple/Microsoft can't match","Whole video is mostly a pitch funnel to Julian's $AI Profit Boardroom community"],"tools":[],"skill_candidates":[],"verdict":"skip","verdict_reason":"Pure product hype/news rehash with no actionable tooling, URLs, or patterns Mike can use; heavily padded with a community upsell."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Julian Goldie SEO Video title: Hermes Agent 3.0 is INSANE! TRANSCRIPT (first 6000 chars): Hermes Agent 3.0 is insane. The AI agent that actually finishes what it starts. What if the AI agent you've been using this whole time was never actually finishing your work? What if every time you stepped away and came back to a task complete message, the agent had quietly given up halfway through? What if the reason your automation keeps failing isn't the model, it's the framework underneath? And what if there's a free open-source agent that just fixed all of that and almost nobody is talking about it? Hey, I'm the digital avatar of Julian Goldie. I help people learn and actually use AI tools in their work. Today, we're breaking down Hermes Agent version 0.13, the Tenacity release. I'm going to walk you through what it is, what's new, how to install it step by step, and my top pro tips for beginners. Let's get into it. So, first, what is Hermes Agent? Hermes Agent is a free open-source self-improving AI agent built by news research. Most agents are session tools. You open them, give them a task, close them. Next time you open them, they've forgotten everything. You start over every single time. Hermes runs the opposite way. It runs persistently. It remembers what it's done. It builds skills from experience and reuses them. And it talks to you through whatever messaging platform you already use, Telegram, Discord, WhatsApp, Slack, and now Google Chat. The core idea is simple. The longer Hermes runs, the better it gets. After a workflow involving five or more tool calls, Hermes automatically saves a reusable skill file. Next time a similar task comes up, it loads that skill and skips the discovery work. It also supports scheduled automations through a built-in cron system, spawns isolated sub aents for parallel work streams, and has full web control, search, extract, browse, vision, image generation, and text to speech. It connects to any MCP server for extended tool capabilities. You can run it on a VPS, a GPU cluster, or serverless infrastructure. It's not tied to your laptop, talk to it from Telegram while it works on a cloud machine. And it works with any model. Now's portal open router with 200 plus models, OpenAI, Anthropic, or your own endpoint. Switch with Hermes model. No code changes, no lockin. Hermes launched February 25th, 2026. By [ASSISTANT] {"tldr":["Hermes Agent v0.13 'Tenacity' ships a durable multi-agent kanban with heartbeats, zombie detection, retry budgets, and auto-blocking on incomplete work — directly targets the 'agent drops tasks silently' problem","Persistent skill files auto-save after any 5+ tool-call workflow and reload on similar tasks, plus Goal Lock keeps objectives anchored across long sessions","Checkpoints v2 adds disk guardrails and autoresume so crashes/restarts don't lose work; cron now has a no-agent watchdog for lightweight scheduled tasks","Talks to Telegram/Discord/Slack/WhatsApp/Google Chat, runs on VPS/GPU/serverless, model-agnostic via OpenRouter (200+ models), MIT licensed with local data","Security wave: 8 CVEs closed, secret redaction on by default, Discord allowlists scoped to guild, SSRF hardened — update immediately if on older version with messaging/MCP"],"tools":[{"name":"Hermes Agent","url":"https://github.com/newsresearch/hermes","description":"Free open-source self-improving persistent AI agent with multi-platform messaging, skill files, cron, and MCP support"}],"skill_candidates":[{"slug":"durable-agent-kanban","description":"Heartbeat + zombie detection + retry budget + auto-block pattern for multi-agent task boards so agents can't silently abandon work"},{"slug":"auto-skill-save-on-5plus-toolcalls","description":"Workflow that detects 5+ tool-call sequences and auto-compiles them into reusable skill files for next-time reuse"},{"slug":"goal-lock-anti-drift","description":"Persistent objective anchoring across long multi-turn sessions to prevent agents drifting off-task"}],"verdict":"dont-miss","verdict_reason":"Hits 4+ Mike domains (AI agents, MCP, Discord/Telegram bots, scheduler/cron, skills, tool-building) and the durable kanban + auto-skill-save patterns are directly applicable to Mike's Carlos/Oliver orchestration and Hermes Mission Control work."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Julian Goldie SEO Video title: OpenClaw 5.16: What You Need To Know… TRANSCRIPT (first 6000 chars): Open 5.16 just showed up on GitHub as a pre-release beta. And there are two things in this update that make your AI agents way more useful. First of all, your agent can now use Grock without an API key. If you have a super Grock subscription, you just log in. No extra cost, no developer portal. Your agent gets full access to Grock through your existing subscription. This is a big thing that's just been released inside Heromus Agent. It's a bit of a game changer. Second, your Telegram agent can now sit inside a group chat, listen to every single message, and stay completely quiet until you tell it to speak. It reads the room. It builds context. And when someone finally asks a question, the answer is 10 times better because it already knows what everyone else has been talking about. I'm going to walk you through everything in this update, why it saves you time, and the one setup that turns all of this into a full AI operating system you control from one screen. Let's get into it. So, let's start with Grock because this one helps you save cost straight away. Right now, if you want your OpenClaw agent to use XAI's Grock, you need an API key. That means going to XAI's developer site, add in a card, buy an API credits, and if you're paying for your Super GRUP subscription, and then paying again for API access, well, that's two bills for the same AI. OpenClaw 5.6 aims to fix that. If you have a Super Grock subscription, you just log in with your account. Your agent uses Grock through your subscription. There's no API key, no extra bills. Just log in and go. Grock has been getting really good, especially for finding up-to-ate information, creating videos, creating images, and now your agent will be able to use all of that through your existing subscription. When this drops, one login and you're done, right? So, this is in the pre-release that's coming from 5.16 inside Open Claw. Now, if you've been paying for Super Grock, but only using it on the website, well, this is a big deal, right? Your OpenClaw agent now has access to everything Grock can do for free because you're already paying for it. Now, the Telegram feature. This one is going to save you a lot of time if you run group chats. Right now, your agent in a Telegram group has two optio [ASSISTANT] {"tldr":["OpenClaw 5.16 pre-release lets agents use Grok via your Super Grok subscription login (no API key, no double billing)","New Telegram 'silent listener' mode: agent reads every message in a group chat but stays quiet until directly asked, building full context first","Scheduled tasks now support blocking with timeout so workflows can chain steps that depend on prior completion","Skill list now cached between messages instead of rebuilt every turn — faster agent response times","Bug fixes: Discord reconnection no longer drops the agent, Matrix encrypted rooms send properly, dashboard shows context window usage"],"tools":[{"name":"OpenClaw 5.16","url":"https://github.com/openclaw","description":"Pre-release beta adding Grok subscription auth and Telegram silent group listener mode"}],"skill_candidates":[{"slug":"telegram-silent-listener-agent","description":"Pattern for building a Telegram group agent that ingests all messages for context but only responds when directly addressed"},{"slug":"blocking-scheduled-tasks","description":"Workflow pattern for chaining scheduled agent tasks with wait-with-timeout so dependent steps fire in order"}],"verdict":"dont-miss","verdict_reason":"Directly hits Mike's OpenClaw/Telegram/scheduler/skills domains with a concrete update he runs in production, plus the silent-listener pattern is a novel agent UX worth stealing."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Julian Goldie SEO Video title: Hermes Agent: Free AI SEO agent is wild… TRANSCRIPT (first 6000 chars): Hermes agent OS is the fastest way to get more customers from Google right now. That means more leads, more people finding you, more sales without a content team, without doing it manually, without any technical experience. And sites using this system have gone from zero to over a thousand new visitors a day from Google. I'll show you proof later. Others are showing up inside chat GPT inside Perplexi and ranking inside Google AI overviews getting in front of customers everywhere people search. And here's what makes this different from every other AI SEO method you've ever seen. The content it creates isn't generic. It's built around your business, your data, your case studies. So it actually ranks because Google rewards original content. And this is as original as it gets. When I show you the system, all you do is you give it a keyword, walk away, come back to published, index, live content, pulling in new customers, and in this video, you're going to see exactly how it works. The whole system live on screen with real results, actual proven case studies and methods showing you how this works. So, stick with me to the end because there's one thing most people skip setting this up when it comes to Hermes agent AI SEO. It's the exact reason their content never actually converts. So, get that part right and everything changes. I'll show you how later. Let's get into it. Today, I'm going to show you how to rank number one with Hermes AI AgentOS. So, an operating system for using SEO. And this is super powerful. And the other cool thing about this is you can use it for free because I'll show you a way to use Hermes for free in a second. So, this is a really powerful system where you can basically generate content and deploy it. Here's a couple of examples of articles that we've created recently targeting our keyword which was Hermes agent OS inside the content. We created a full blog post as you see right here. It's like beautifully organized setup for conversions. And basically what I do is I give it a keyword. I tell it what I want it to create the actual AI agent and using this operating system it will go off and start creating the content directly for me. Basically, you see here it's added internal links, external links, and created [ASSISTANT] { "tldr": [ "Julian Goldie pitches 'Hermes Agent OS' as a Claude-built SEO agent system that takes a keyword, generates branded content with case studies, deploys to your site, and ranks in Google + AI Overviews + Perplexity + ChatGPT.", "Claims case study sites going 0 to 1,134 clicks/day and outranking Hermes' official site for the target keyword using agentic SEO + uniqueness via your own data/case studies.", "The 'system' is essentially Claude + Open Claw + a mission control dashboard managing skills (goals, journal, memory) and SEO content agents — built with Claude itself, not a standalone product.", "Real takeaway: gated lead-magnet funnel for his 'AI Profit Boardroom' — the actual prompts/setup are behind a paid community, not given in the video.", "Naming collision risk for Mike: Julian's 'Hermes Agent OS' has nothing to do with Mike's Hermes Mission Control / MCC project." ], "tools": [], "skill_candidates": [], "verdict": "skip", "verdict_reason": "Pure lead-magnet promo with no novel pattern, no real tool URLs, and no technique Mike isn't already doing 10x deeper with his own SEO agent fleet (Einstein/Tommy/Ghost/Shakespeare) and Hermes MCC." }

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Income stream surfers Video title: RIP: Claude Opus 4.7 + Claude Code Just Had a WORLDWIDE Outage TRANSCRIPT (first 6000 chars): Hey guys, welcome to this video. Going to be a quick one today, but basically Claude is down again, as you can see right here. It looks like it's just not working at all, at least for me. Anyway, I'm getting this error that you can see right here. Yes, I'm playing RuneScape at the same time. But uh API error 500 internal error. This is a server side issue, usually temporary. Try again in a moment. If it persists, check claude.us status.claude.com. So, I went here and I had a little look and apparently it was also down on the 15th. Error elevated error rates on requests to some models. So, basically guys, I would say as an alternative, you have the Codex uh app. So, yeah, basically guys, I thought I would talk basically about the alternatives that we have in today's video. So, I'm going to do something a little bit meta here. I'm actually going to use um Codex to create a thumbnail for this video. So, Claude is down right now. Can you use my thumbnail thumbnail skill to make a thumbnail about the fact that Claude is down. Uh worldwide outage. Okay, so yeah, let's talk about the other alternatives that you have, right? If you don't want to use Codex, I would say Codex is your number one. Right, so this is 5.5. I use I would recommend using this on low just because it does take a hell of a long time if you're not on low, right? So, we have we cannot use Opus 4.7 right now. Okay, so Kimmy 2.6, definitely a good alternative um inside Kimmy code. Again, I'm not talking about like is this as good as Opus 4.7? No, it's not. But, if you're going to be using uh a model today and you actually want to get something done, Kimmy code, definitely a good alternative. These models here, Deep Seek, Quen 3.6, Mimo inside Open code would also be a very very good way to do this. Open code is free open source to use. All you have to do is pay for the gas the the AI models basically, right? The benzina, the petrol. I forgot the word in my head then, so I only knew the Italian word. Looks like they actually have a desktop app now, which is really really nice. I haven't used this yet, but I will definitely give this a test. Let's actually just download this um download this now. So, I'm on Apple silicon, I think. So, there we go. [ASSISTANT] {"tldr":["Claude Opus 4.7 + Claude Code had a worldwide outage on 2026-05-17 — check status.claude.com when 500 errors hit","Primary fallback: Codex app with GPT 5.5 on low effort ($20 ChatGPT plan is enough for most coding)","Budget fallback inside Open Code: Mimo V2.5 (Xiaomi) is the sleeper pick, with Qwen 3.6 27B and DeepSeek as backups","Kimmy 2.6 inside Kimmy Code is a credible secondary if you want a Claude-feel alternative","Antigravity + Gemini 3 Flash is a free-tier option Google has basically abandoned but still works"],"tools":[{"name":"Claude Status","url":"https://status.claude.com","description":"Official Anthropic status page to check Claude/Claude Code outages"},{"name":"Codex","url":"https://chatgpt.com/codex","description":"OpenAI's coding agent app — GPT 5.5 on low/high effort, included with $20 ChatGPT plan"},{"name":"Open Code","url":"https://opencode.ai","description":"Free open-source coding agent — bring your own model via OpenRouter (Mimo, Qwen, DeepSeek)"}],"skill_candidates":[{"slug":"coding-agent-failover","description":"When Claude Code is down or nerfed, route to ranked fallback: Codex (GPT 5.5 low) → Kimmy Code → Open Code + Mimo V2.5 → Antigravity + Gemini 3 Flash. Includes status-check probe and model effort tuning."}],"verdict":"worth-a-skim","verdict_reason":"Timely outage commentary with a useful failover ladder (Codex → Kimmy → Mimo/Qwen via Open Code) but no new tools Mike doesn't already know and no novel patterns beyond 'switch to Codex when Claude breaks.'"}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: In The World of AI Video title: Stop Paying $200/Month for Claude Code! Here's a FREE Option! TRANSCRIPT (first 6000 chars): We all know claude code and its powerful agent capabilities right inside your terminal. With all the recent hype around it, many developers have been searching for cheaper or even free alternatives that delivers similar results. Today, I found one that's truly impressive. Aentic efficient in code generation that's open- source and partially free. It's from the Mistrol team who's also sponsoring today's video and it's called Mistral Vibe. This is something that launched recently and it's already shaping up to be a gamecher in AI assisted development. But what is Misal Vibe? Well, it's a terminal native AI coding agent that is like open code and cloud code capable of writing, testing, refactoring your code, and even deploying your code all from the command line. It supports async tasks. It integrates with tools like GitHub, Jira, Slack, and Genkins, and comes with customizable agent workflows to automate long running tasks. For example, take a look at how it can document shell aliases with the Mistro 5 tool. This is where it's capable of reading your DOT files, understanding what each alias does, and generating clean structure documentation automatically. You can point it at your aliases and explain each command purpose. It filters out highly specific scripts and shows how to install them via the Zshrc or the alias file. You can even create custom sub agents in Mistral Vibe, letting you build specialized agents to deploy scripts, PR reviews, test generations, or even repeatable workflows and invoke them on demand with a single command. Sub agents inherit project context while focusing on your domain. It is something that can enable composable automation that scales with your team and your developer workflow. You even have multi-choice clarification that gives you control when Vibe isn't sure what to do next. Instead of guessing, it surfaces multiple options so you can pick the path that matches your intent. This means safer automation and sharper outcomes. And finally with slash command skills. This makes repeated tasks faster because with mistrol vibe CLI it supports preconfigured workflows that you can access with a single keystroke. You simply use the slash command to deploy lint generate docs or even have it [ASSISTANT] {"tldr":["Mistral launched Mistral Vibe, a free open-source terminal-native AI coding agent positioned as a Claude Code alternative","Ships with sub-agents, slash command skills, multi-choice clarification prompts, and integrations for GitHub/Jira/Slack/Jenkins","Powered by Devstral (codegen), Codestral (context), and Codestral Embed (semantic search) — bring your own API key or use free tier","Paid Le Chat Pro at $14.99/mo unlocks higher usage; Team plan $24.99/seat; 50% student discount","Install via curl/bash or Python, plus VS Code and JetBrains extensions with tab completion"],"tools":[{"name":"Mistral Vibe CLI","url":"https://chat.mistral.ai","description":"Free open-source terminal AI coding agent from Mistral, Claude Code alternative with sub-agents and slash skills"}],"skill_candidates":[{"slug":"mistral-vibe-cli","description":"Install, auth, and operate Mistral Vibe CLI as a fallback/secondary coding agent alongside Claude Code with sub-agents and slash skills"}],"verdict":"worth-a-skim","verdict_reason":"New Claude Code competitor worth knowing about for agentic coding lane, but it's a tool announcement video with no novel patterns and Mike already runs Claude Code + Codex as his primary coding agents."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Eric W Tech Video title: How I Convert NotebookLM Slides into Editable PPT with OCR | NotebookLM + PDNob AI TRANSCRIPT (first 6000 chars): You just spent time building a clean slide deck from your research in Notebook LM. The sources are organized, the points make sense, the structure is ready. Then you export or save that work as a PDF, open it later, and realize the real problem. It looks like a presentation, but it does not behave like one. You try to click on the title, nothing. You try to select a paragraph, it selects the whole page, or it does not select cleanly. You try copying the text into PowerPoint and the formatting falls apart. Images are locked in place, bullet points lose their spacing. Now, instead of editing a presentation, you are basically rebuilding the whole thing from scratch. That is the workflow we are fixing today. In this video, I will show you how to take a Notebook LM PDF slide file and convert it into an editable PowerPoint using Tenorshare PD Nob 2.0. We will go step-by-step. Open the PDF, run OCR if the content is locked, convert it to PowerPoint, and then check the final PPT file so you can see what is editable. Let me show you what I mean. Here is a Notebook LM PDF I exported earlier. This is a simple slide deck based on research notes. It looks fine when I scroll through it. The layout is clean, the titles are readable, the images and sections are already in place. But the problem starts when I actually need to change something. Watch what happens when I try to edit this title. I cannot just click into it like a normal PowerPoint text box. And if I try to copy the slide content manually, I am probably going to lose the formatting. That means more cleanup, more resizing, and more time spent doing work that should already be done. So the goal is simple. I want this PDF to become a real PowerPoint file. I want text I can edit, I want images I can move. That is where PD Nob comes in. This is Tenorshare PD Nob 2.0. It is a PDF editor built for editing, converting, OCR, compression, annotation, and working with PDFs in a more practical way. For this workflow, the two features that matter most are OCR and PDF to PowerPoint conversion. OCR is what helps when the PDF content is image based or scanned. It detects the text inside the PDF and turns it into something the software can work with. Then the convers [ASSISTANT] {"tldr":["Tenorshare PDNob 2.0 converts NotebookLM-exported PDF slide decks into editable PowerPoint files via OCR + PDF-to-PPTX export","Workflow: open PDF in PDNob, run OCR if text is image-locked, convert to PPTX, then edit in PowerPoint instead of rebuilding from scratch","OCR step is the unlock — it's what makes locked NotebookLM slide text selectable and editable post-conversion","Layout/slide order/images are mostly preserved, so cleanup is minor vs starting over","Affiliate-style walkthrough for a single commercial PDF editor — not a novel pattern or agent workflow"],"tools":[{"name":"Tenorshare PDNob 2.0","url":"https://pdnob.tenorshare.com","description":"Commercial PDF editor with OCR and PDF-to-PowerPoint conversion"},{"name":"NotebookLM","url":"https://notebooklm.google.com","description":"Google's research/notes tool that can export slide-style PDFs"}],"skill_candidates":[],"verdict":"skip","verdict_reason":"Sponsored walkthrough of a consumer PDF-to-PPT converter — zero overlap with Mike's agent/SEO/automation domains and no reusable pattern worth extracting."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: DIY Smart Code Video title: Anthropic just shifted the entire AWS landscape #ai #aws #trending TRANSCRIPT (first 6000 chars): Every AWS dev shipping with Claude knows this pain. Bedrock gets the new features a week late. But today, that leg is gone. Anthropic just shipped the full Claude platform on AWS. Generally available today. And here's the part most people will miss. This is not Bedrock with a new name. Bedrock is the data resident subset. Platform on AWS is the full native ASPI. Day zero. Which means the entire beta service lands on AWS today. Managed agents, advisors, skills, files, MCP connector. Code execution, web search, and the Claude console on AWS for the first time. The prompt improver, the generator, the evils UE. And here's how it ships through your AWS account. The endpoint runs on AWS external anthropic.region.api.aws. Auth is through identity access management. Every call is a CloudTrail event. Billing rides on AWS marketplace, retiring against your existing commitment. Plus, it's available in 18 AWS regions at GA. That's broader than most Claude routes at launch. Three named customers are already shipping on it. ReliaQuest in cybersecurity. Open router routing through AWS identity access management. And Emergent with day one access to every new model. So, when do you pick which? If you want every beta from day one, that's platform on AWS. If your data must stay inside AWS, that's Bedrock. The version leg is over. Is this the end of Bedrock for new agents? Tell me in the comments. Subscribe for more AI news. If you want to learn more about AI, check out the dynamis.ai community. Extract the structured JSON signal per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] {"tldr":["Anthropic shipped the full Claude platform natively on AWS (not just Bedrock) — every beta feature lands day zero","Includes managed agents, Skills, Files, MCP connector, code execution, web search, and the Claude console on AWS","Auth via IAM, logging via CloudTrail, billing via AWS Marketplace (burns down existing AWS commit)","Available in 18 AWS regions at GA — broader than most Claude launches","Decision rule: use Platform on AWS for day-zero features, stick with Bedrock only if data must stay inside AWS"],"tools":[{"name":"Claude Platform on AWS","url":"https://aws.amazon.com/marketplace","description":"Full native Anthropic API on AWS with day-zero feature parity, IAM auth, CloudTrail logging"},{"name":"Dynamis.ai","url":"https://dynamis.ai","description":"AI learning community plugged at end of video"}],"skill_candidates":[],"verdict":"worth-a-skim","verdict_reason":"Relevant AWS/Claude news with a clean decision rule (Platform vs Bedrock), but Mike runs on Vercel/Supabase not AWS, so no immediate action — just useful context if a client ever demands AWS-resident Claude."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: DIY Smart Code Video title: Ripgrep Just Made Claude Code Search 10x Faster #ClaudeCode #Shorts TRANSCRIPT (first 6000 chars): 14 features, 80 fixes, 20 improvements, four Claude code releases dropped in 4 days. Here is what actually shipped. First, the built-in grep tool is now Ripgrep. Searching your code base just got faster, more accurate, and consistent across every operating system. Second, the Claude agents command picks up a serious flag pack. Add dir, settings, MCP config, plugin dir, and permission mode. Pin a sub agent to exactly the tools and rules you want. Third, the plugin marketplace finally shows per turn token cost estimates. You see the price tag before you install. No more guessing what a plugin will cost you. Fourth, the rewind menu adds summarize up to here. Compress the old turns, keep the recent ones. Long sessions stop bleeding tokens. Fifth, hooks now emit terminal sequences directly from JSON output. Notifications, window titles, even bell acts. No TTY required. Sixth, background sessions finally survive a Mac going to sleep. The demon catches clock jumps on wake, and your model and effort level persist across the resume. Ripgrep, Claude agents flags, or plugin price tags. Which one wins? Drop a comment. If you want to learn more about AI, check out the dynamis.ai community. Extract the structured JSON signal per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] {"tldr":["Claude Code's built-in grep is now Ripgrep — faster, more accurate, OS-consistent search across your codebase","claude agents command gained --add-dir, --settings, --mcp-config, --plugin-dir, and --permission-mode flags to pin subagents to exact tools/rules","Plugin marketplace now shows per-turn token cost estimates before install — no more blind plugin spend","Rewind menu adds 'summarize up to here' to compress old turns and stop long sessions from bleeding tokens","Background sessions survive Mac sleep, hooks emit terminal sequences from JSON output (no TTY needed)"],"tools":[{"name":"Ripgrep","url":"https://github.com/BurntSushi/ripgrep","description":"Now the default grep engine inside Claude Code for faster cross-OS code search"}],"skill_candidates":[{"slug":"claude-agents-pinned-config","description":"Pattern for pinning subagents to exact tool/rule sets using the new claude agents flag pack (--add-dir, --settings, --mcp-config, --plugin-dir, --permission-mode)"},{"slug":"claude-code-token-budget-hygiene","description":"Use rewind 'summarize up to here' plus plugin marketplace cost previews to keep long sessions and plugin loadouts under a token budget"}],"verdict":"dont-miss","verdict_reason":"Hits Mike's Claude Code + skills + MCP domains directly with shipped features he uses daily — subagent flag pinning and rewind summarize materially change how Oliver/Carlos sessions get configured."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: DIY Smart Code Video title: Mike Krieger Could Build Instagram Solo Now - The Founders Playbook 2026! TRANSCRIPT (first 6000 chars): Anthropic just published a 35-page playbook, and the first warning is that Claude code will help you fail fast. Page 10. 42% of startups failed building something nobody wanted. And the playbook says that number is only going to climb. Which is why Mike Krieger, Instagram's co-founder, said publicly he could have built Instagram with just himself and his co-founder using Claude today. Krieger said it. Anthropic just made it operational. So, Anthropic remapped startup building into four stages. Idea. MVP, launch, scale. Each one has a goal, an exit test, and a failure mode. Idea is finding a problem worth paying for. MVP is proving people actually will. Launch is going from one founder to a small team. And scale is when every decision still going through, you stops being an asset and starts being the bottleneck. Each stage has a named killer. Idea stage. Mistaken building for validating. You ship a prototype in 4 hours and call it proof. MVP stage. Agentic technical debt. Skip the specs, and every Claude session drifts the code base further from your plan. Launch stage. The founder becomes the bottleneck. Decisions that should take an hour now take a week. Scale stage. Delegating the operational layer. You built the systems, and now you have to actually trust them. Four stages. Four named ways to die. So, which Claude do you reach for first? Plus, page 11 publishes a decision table. Three services, same brain, different jobs. Chat for quick exchanges, Claude co-work for knowledge work. Claude code for the code base. Pick wrong, you waste hours. Here's what running it actually looks like. Welkin Technologies, 4 months, three founders. One technical. They beat the consulting firms to a state contract. $11 million seed. The discipline works. Skip the cloud.md file and every cloud code session re-derives your decisions from scratch. The playbook calls 5 minutes of docs per session cheap insurance against a code base that collapses. So, here's the question. Are you running this playbook or still validating with wipes? Drop your stage in the comments. And subscribe for more AI news. Extract the structured JSON signal per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] {"tldr":["Anthropic published a 35-page founder's playbook mapping startup building into 4 stages (Idea, MVP, Launch, Scale) — each with a named failure mode worth memorizing","Page 11 has a decision table for which Claude product to use when: Chat for quick exchanges, Claude co-work for knowledge work, Claude Code for the codebase","Skip the CLAUDE.md file and every Claude Code session re-derives decisions from scratch — playbook calls 5 min of docs per session 'cheap insurance' against agentic technical debt","Welkin Technologies proof point: 3 founders, 4 months, beat consulting firms to a state contract and raised $11M seed using this discipline","Named killers per stage: idea=mistaking building for validating, MVP=agentic tech debt, launch=founder bottleneck, scale=failing to trust delegated systems"],"tools":[{"name":"Anthropic Founder's Playbook 2026","url":"https://www.anthropic.com","description":"35-page playbook on building startups with Claude across 4 stages"},{"name":"Claude Code","url":"https://claude.com/claude-code","description":"Anthropic's agentic coding tool used as the codebase brain in the playbook"}],"skill_candidates":[{"slug":"claude-md-session-discipline","description":"Enforce 5-minute CLAUDE.md doc update per Claude Code session to prevent codebase drift and agentic tech debt across long-running projects"},{"slug":"founder-stage-diagnostic","description":"Diagnose which of the 4 startup stages (Idea/MVP/Launch/Scale) a project is in, identify the stage-specific killer, and route to the right Claude product (Chat/co-work/Code)"}],"verdict":"worth-a-skim","verdict_reason":"Surface-level summary of an Anthropic playbook Mike should read directly — the source doc is the real signal, the video is just a pointer with one useful reminder (CLAUDE.md hygiene) that Mike already practices."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: DIY Smart Code Video title: 900 stars in 24 hours — what's the catch with Zero? #vercel #programming TRANSCRIPT (first 6000 chars): Versal Labs just dropped a new programming language. It is called Zero and it is built for agents, not humans. Chris Tate posted this on May 16th. By the next morning, the repo had over 900 stars. The tagline, the programming language for agents, a systems language for small native tools, explicit effects, predictable memory, and structured compiler output. Here is what a hello world looks like. Notice the file extension. dot zero numeric zero inside main. You receive a world capability. If you want to write to standard output, you have to accept that world. The function declares races, meaning it can fail. And every side effect is prefixed with check. So the compiler can see exactly where you touched the outside. Now look at this build line from the announcement. 16 kilobytes, 1 millisecond. That is the entire hello binary statically linked. No runtime. Almost every command can emit JSON. Zero check, zero run, zero graph, zero size. Zero roots, zero doctor. All of them speak structured output because the audience is not a human reading a terminal. The audience is an agent reading a pipe. Now the honest part, zero is experimental. The Redmi says it directly. The language is not stable. There is no public model system documented yet. No published benchmarks. The example programs are all under 50 lines. And the compiler is a brand new C code base with a self-hosted rewrite already in progress. So here is the real question. Is this a genuine new category? A systems language built from day zero for the way agents actually write code or is it just go with better error JSON in aversal hoodie? Drop your verdict below. Real language or hype with a dot. Zero extension. I want to read your take. Extract the structured JSON signal per your instructions. Return ONLY valid JSON, no markdown fences. [ASSISTANT] {"tldr":["Vercel Labs dropped 'Zero' — a systems language designed for agents (not humans) with explicit effects, capabilities, and structured JSON output from every CLI command","Hello world compiles to a 16KB statically-linked binary in 1ms with no runtime — every side effect is prefixed with 'check' so the compiler tracks exactly where you touch the outside world","Repo hit 900+ GitHub stars in 24 hours (posted May 16 by Chris Tate) but is explicitly experimental: no stable spec, no module system, no benchmarks, compiler is brand new C","Every command (zero check/run/graph/size/roots/doctor) emits JSON because the consumer is an agent reading a pipe, not a human reading a terminal","Open question: genuine new agent-native language category, or Go with better error JSON in a Vercel hoodie"],"tools":[{"name":"Zero","url":"https://github.com/vercel-labs/zero","description":"Vercel Labs experimental systems language built for AI agents with explicit effects, capability-passing, and structured JSON compiler output"}],"skill_candidates":[{"slug":"agent-native-cli-output-spec","description":"Pattern for designing CLI tools where every command emits structured JSON by default so agents can consume output without parsing terminal text"}],"verdict":"worth-a-skim","verdict_reason":"Hits Mike's Vercel + agentic-coding + tool-building domains and the agent-native CLI design philosophy reinforces his CLI-first routing rule, but Zero itself is too experimental to use today — bookmark the pattern, not the language."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Cole Medin Video title: Pi is INCREDIBLE - Building a Custom Coding Agent Live TRANSCRIPT (first 6000 chars): Time to dive into Pi. This is a coding agent. It's been around for a while now, like at least a few months, but I've just started to like really really pay attention to it, especially because of how bad the rate limits have been for other coding agents right now. And so, the idea behind Pi is it is a minimal coding agent. It's made to be a coding agent that you build on top of instead of taking a really massive bloated tool and trying to retrofit it to how you like to code. So the idea here is that we adapt PI to our workflows instead of the other way around. And if you've been following my channel at all, you know that uh this is really in line with my ethos, right? Like I am generally a fan of building your own systems because then you can really take control of it and mold it and customize it to how you like to work. And so a lot of the work that I've been doing with Archon and just generally teaching on harness engineering recently, it fits really really well with Pi. So honestly, it's kind of crazy that I haven't covered this more until now. So I did do one video on my channel where I used Pi with Archon and I showed the integration that I built with Pi and Archon. So, if you look at the providers that we support in Archon, let me just go down in the read me here. Um, goodness, there we go. Okay, so for AI assistant clients, we support Claude, Codeex, and Pi. So, Archon is my open source tool that helps you build your own harnesses. It's very in line with PI because both are all about, you know, create your own process, build your own workflows and trying instead of trying to just use something off the shelf that can be pretty hard to work with. Uh, and so yeah, I'm going to do quite a few things with Pi in our live stream today. And I'm gonna keep the stream pretty casual overall. I think one thing with the live streams I've been doing the past couple of weeks is I've been trying too hard to like explain everything really thoroughly as I'm doing it. And so it makes the stream move kind of slow. So I think my long form, you know, YouTube content, the videos I put out, that's where I should really be explaining things. But here, I just want to build live with you. So, we're going to explore Pi. I've obviously done quite [ASSISTANT] {"tldr":["Pi is a minimal coding agent harness you build ON TOP OF instead of retrofitting bloated tools — perfect fit for Mike's 'minimalist engineer' ethos","Pi supports tons of providers out of the box (Kimi, Qwen via OpenRouter, Codex sub, Copilot) — no workarounds needed unlike Claude Code","Pi has an extension marketplace + lets you build custom extensions (subagents, MCP adapter aren't native — you install/build them)","Cole integrated Pi with Archon (his open-source harness builder) alongside Claude and Codex","Pi pairs naturally with self-evolving codebase patterns like Cole's Dark Factory project"],"tools":[{"name":"Pi","url":"https://github.com/PiCoder","description":"Minimal coding agent harness built to be extended, not retrofitted"},{"name":"Archon","url":"https://github.com/coleam00/Archon","description":"Cole Medin's open-source harness builder — supports Claude, Codex, Pi"}],"skill_candidates":[{"slug":"pi-coding-agent-setup","description":"Install and configure Pi with provider routing (Kimi, Qwen via OpenRouter, Codex sub) plus extension marketplace install workflow"},{"slug":"build-custom-pi-extension","description":"Pattern for authoring a custom Pi extension (subagent, MCP adapter, or workflow) and publishing to the Pi package catalog"}],"verdict":"dont-miss","verdict_reason":"Hits multiple Mike domains (agentic coding, LLM tooling, tool-building, harness engineering) and introduces a minimal-by-design coding agent Mike hasn't covered that directly aligns with his 'build your own tools, simplicity over cleverness' trait."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: Clearmud Video title: Install Hermes Agent on Windows in 10 Minutes (COPY ME) TRANSCRIPT (first 6000 chars): In this video, I'm going to walk you through the steps it takes to install Hermes agent onto a Windows-based machine. So, without further ado, let's build. All right, so as you can see here on the screen, here is the main website. It's hermes-agent.newsresearch.com. And as you can see here, this is the command if you want to install it on any Mac OS or Linux-based distribution. However, we're on Windows, so we want to click this WSL2 to install it on Windows. And this is going to give us instructions on how to prep Windows to allow that install to occur. So, I already have PowerShell on my bar. Alternatively, if you don't, you can just press start, search PowerShell. And what you want to do is right click, select run as administrator. All right, now for those curious, I know I'm on Windows 10, I just prefer it. These same exact steps will work on Windows 11. So, what we want to do is type in WS--install. Click yes. Okay, Ubuntu has been installed. Now, we do have to reboot the system. So, I'm just going to go ahead and restart this machine. All right, so what we want to do is select start menu, and as you can see here, we do have Ubuntu installed. We can just select there. And let's go ahead and open up the browser one more time. Go back to News Research's website here, and we want to grab this curl command. We're just going to copy that there. And let's go into Ubuntu. And this may take a few minutes as it says here on the screen. Let's just wait patiently. All right, so as you can see here, looks like it is installed. Now, we do need to enter in a username for the server. So, for the sake of the demo, I'm just calling this Hermes demo. And we're going to add in a password here. Wonderful. So, installation is successful now. Perfect. So, now what we want to do is paste in Looks like this copy command doesn't want to work natively here. So, let's just go ahead and type it in manually here, shall we? curl https://hermesagent. newsresearch.com/install.sh and bash. And here we go. I'm guessing it's because I have this actually we're accessing a Windows 10 PC via remote desktop, so my keybinds aren't working properly. And sure, why not? Now, we're going to enter in our password that we created. Perfect. So, now it's taking us thr [ASSISTANT] {"tldr":["Install Hermes Agent on Windows via WSL2 + Ubuntu, then curl the install.sh from hermes-agent.newsresearch.com","Quick setup wizard auths against OpenAI Codex (or other model providers) using device code flow","Demo connects GPT-5.5 as the controlling model — creator claims night/day difference vs 5.4 for agent work","Agent runs locally in terminal via `hermes` command after shell reload — no cloud dashboard required","No actual agent build shown — install walkthrough only, follow-up videos teased for use cases"],"tools":[{"name":"Hermes Agent","url":"https://hermes-agent.newsresearch.com","description":"Terminal-based AI agent installable on Mac/Linux/WSL2 with pluggable model backends including OpenAI Codex"}],"skill_candidates":[],"verdict":"skip","verdict_reason":"Surface-level install walkthrough for a competing terminal agent — Mike already runs Claude Code + Codex with a full orchestration layer, no new pattern or tool worth adopting here."}

[USER] You are an expert content analyst for Mike Merlino, an AI agency operator and builder. Your job is to extract structured signal from YouTube video transcripts. Mike's domains of interest (if a video hits 2+ of these → strong signal): [ "AI agents", "LLM tooling", "Claude Code", "MCP servers", "skills", "prompt engineering", "SEO automation", "GMB", "local SEO", "cold outreach", "SMS", "GoHighLevel", "Next.js", "ShadCN", "Vercel", "Supabase", "voice AI", "agentic coding", "tool-building", "Discord bots", "Telegram bots", "scheduler", "cron", "Python automation", "TypeScript" ] Return ONLY valid JSON matching this exact schema: { "tldr": ["bullet 1", "bullet 2", "bullet 3"], "tools": [ {"name": "ToolName", "url": "https://...", "description": "one line"} ], "skill_candidates": [ {"slug": "kebab-case-name", "description": "what skill this would capture"} ], "verdict": "dont-miss" | "worth-a-skim" | "skip", "verdict_reason": "one sentence explanation" } Verdict rules: - "dont-miss": new tool + active use-case for Mike OR novel pattern OR hits 2+ Mike domains OR introduces something Mike hasn't seen - "worth-a-skim": solid content, 1 Mike domain, no new tools - "skip": opinion/news/rehash with no actionable takeaway Be OPINIONATED. Do NOT default to neutral. Mike trusts your judgment. TL;DR bullets: 3-5, action-oriented, specific. No vague summaries. Tools: only real URLs you are confident about from transcript context. Omit if uncertain. Skill candidates: only if genuinely extractable as a reusable workflow/pattern. Channel: All About AI Video title: Find 100x Low Risk High Reward Polymarket Strategies With AI TRANSCRIPT (first 6000 chars): Hello, hope you are doing well. So, we are back again on Poly Market. I want to do some trading. I want to look up some strategies using codeex cloud code and you saw the title, right? And just to show you that it's not clickbait. You can kind of see the trade I did here, right? You can can we zoom in a bit? So, you can see the odds were 1%, right? We put one up, we went we 100, right? So, that's a 100x trade. So, this is a strategy that's a bit strange. So, I kind of wanted to kind of look a bit more into it today, maybe using like claw code and stuff and just to see what happens in these strange trades. You can see we have a bunch of these. We have this, for example, this one. This is kind of on the Bitcoin up and down. How do I get this? Okay, so you can see again this time we have the 2% there and we've been 50 from one, right? So, this is a bit of a strange trade. So, I'm going to try to look a bit into that today and kind of see how I did this. Okay. So, if we go to cloud code here now, you can kind of see uh this is just confirmed these trades, right? Just a proof of trade. So, this time we got the 48x here, right? We put up one, we got 59 back. And we can of course confirm this by going to kind of the poly scans here if you wanted to just to see that it's actually going through and it's just that not just like a visual bug. So you can see this was success and yeah you can go back and check this if you want to but basically the idea that I wanted to do in this video is basically just to look at these trades and to see exactly what is happening because they are very strange and this kind of like a low risk high value but again it's not really so it's a bit strange but it's really fun to use like claw code or like uh codeex to analyze these trades. and try to learn from them and make it into like a repeatable strategy. So I thought just talk a bit about that today and I'll show you my setup here. So the way I stumbled over this trade was I think it was just by accident. So I kind of wanted to dive more into it. So if we just copy all of this kind of this proof of trade we extracted and let's say you start up like a new codeex. Uh I want to do like dash dash yolo right for codeex. So I want to run anything like d [ASSISTANT] { "tldr": [ "Creator uses Codex/Claude Code in YOLO mode to forensically analyze a 100x Polymarket trade by pasting trade hashes and letting the agent investigate wallets and PolyScan data", "The 'strategy' is exploiting Polymarket's close settlement delay: bot posts resting low bids on both YES/NO sides, gets filled on stale prices after the window closes, guaranteeing payout on the winning side", "Workflow pattern: paste raw transaction proof into agent, ask 'investigate and explain', then iterate with 'brainstorm strats to replicate' — using AI agent as a trade post-mortem analyst", "No code or repo shown — purely conversational analysis of on-chain trade data using --dangerously-skip-permissions flag" ], "tools": [ {"name": "Polymarket", "url": "https://polymarket.com", "description": "Prediction market platform where the trades occurred"}, {"name": "Codex CLI", "url": "https://github.com/openai/codex", "description": "OpenAI's coding agent CLI used in --yolo mode for trade analysis"} ], "skill_candidates": [ {"slug": "trade-postmortem-analyzer", "description": "Paste transaction hashes/proof into an agent and have it investigate on-chain data, explain what happened, and brainstorm replicable strategies — generalizable to any on-chain forensic analysis workflow"} ], "verdict": "skip", "verdict_reason": "Niche Polymarket gambling content with no novel tooling, no actionable pattern for Mike's domains, and the 'strategy' is a one-off exploit of settlement delay that isn't a reusable workflow." }

[USER] Resume Lane A backfill check + commit work Resume Lane A backfill monitoring [ASSISTANT] Both wakeup tasks already done before they fired: - **Backfill: COMPLETE** — 15,201 Pieces rows + 142 pieces_workstream_summary landed in `agent_memories` (total 17,554) - **Commits: DONE** — `pieces-integration 7f9077d` (Lane A patches) + `d882ecb` (backfill truncation fixes) + `master-brain 3153c89` (Lane B + D scripts), all local Nothing to monitor. Awaiting your call on the 3 options from before. <options> <option>Push all 3 commits to GitHub now</option> <option>Done, closing terminals</option> <option>One more lane / task I missed</option> </options>

[USER] <task-notification> <task-id>bxi59ek96</task-id> <tool-use-id>toolu_01K3xqeeyhfkTPjydCtHVb89</tool-use-id> <output-file>C:\Users\mikem\AppData\Local\Temp\claude\D--ClaudeDev-00-GITHUB--working-on-Tools-master-brain\7d0b264a-31bf-4314-9019-518014f87d1d\tasks\bxi59ek96.output</output-file> <status>completed</status> <summary>Background command "Locate merlino-sops MCP categories list" completed (exit code 0)</summary> </task-notification> [ASSISTANT] Standing by for your call on the 3 options. <options> <option>Push all 3 commits to GitHub now</option> <option>Done, closing terminals</option> <option>One more lane / task I missed</option> </options>

[USER] Merlino Marketing Podbean I am just promoting Ranking Reels [ASSISTANT] Got it. Merlino Marketing is the channel, RankingReels is the content. Episode titles are clean, that's what matters. No new Podbean show needed. 44 Podbean live. Rest waits on you. <options> <option>I'll re-auth SoundCloud now</option> <option>I'll contact Zernio support now</option> <option>Wrap and save handoff — pick this up tomorrow</option> <option>Show me the POWD page</option> </options>

[USER] Fire the other 43 on Podbean now (only working channel) [ASSISTANT] ## DELIVERED — 44/44 Podbean episodes live All 44 published, retitled with " | RankingReels" suffix, zero "Merlino Marketing" strings remaining anywhere in titles or descriptions. **Top 10 live URLs:** 1. https://MerlinoMarketing.podbean.com/e/geo-seo-explained-how-to-rank-in-ai-search-chatgpt-perplexity-google-ai/ 2. https://MerlinoMarketing.podbean.com/e/how-businesses-in-columbus-rank-on-google-maps-with-local-seo/ 3. https://MerlinoMarketing.podbean.com/e/how-businesses-in-orlando-rank-on-google-maps-with-local-seo/ 4. https://MerlinoMarketing.podbean.com/e/how-businesses-in-minneapolis-rank-on-google-maps-with-local-seo/ 5. https://MerlinoMarketing.podbean.com/e/how-businesses-in-sacramento-rank-on-google-maps-with-local-seo/ 6. https://MerlinoMarketing.podbean.com/e/how-businesses-in-fort-worth-rank-on-google-maps-with-local-seo/ 7. https://MerlinoMarketing.podbean.com/e/how-businesses-in-san-antonio-rank-on-google-maps-with-local-seo/ 8. https://MerlinoMarketing.podbean.com/e/how-businesses-in-jacksonville-rank-on-google-maps-with-local-seo/ 9. https://MerlinoMarketing.podbean.com/e/how-businesses-in-seattle-rank-on-google-maps-with-local-seo/ 10. https://MerlinoMarketing.podbean.com/e/how-businesses-in-dallas-rank-on-google-maps-with-local-seo/ **POWD updated:** https://powd.rankingreels.com/rankingreels/2026-05-16-bmm-44-drafts/ (pw: `bigmike`) **VERIFICATION:** VISUAL — screenshot at `.planning/proof/podbean-live-2026-05-16.png`. NYC episode HTTP 200 confirmed. **QUEEN:** 9/10 — all 44 live, retitled, manifest + POWD updated. ## Heads up - The Podbean **show name** itself is still "Merlino Marketing" — that's the channel-level branding, not controllable per episode. URLs all stem from `MerlinoMarketing.podbean.com/...`. The episode titles (what listeners see in directories like Apple/Spotify) are clean. - If you want a real RankingReels-branded show, you'd need a new Podbean show created in their dashboard, then migrate or republish. Manual one-time step. ## Still blocked (unchanged) - BMM 11 platforms × 44 + Pinterest 44 → waiting on Zernio worker fix - SoundCloud 44 → waiting on OAuth re-auth <options> <option>Create a new RankingReels Podbean show + republish</option> <option>Move on — Podbean is good enough as-is</option> <option>I'll re-auth SoundCloud now</option> <option>I'll contact Zernio support now</option> <option>Wrap and save handoff</option> </options>

[USER] I see the nav but it says it test data 09:27:04 PM Mock data active. Set NEXT_PUBLIC_FLEET_API + FLEET_BROKER_SECRET to connect to a live broker. [ASSISTANT] **FIXED.** Mock banner gone on all 3 pages. **Root cause:** `message-feed-view.tsx` and `broadcast-view.tsx` accepted the `isMock` prop but rendered the banner unconditionally. Only `peer-grid-view.tsx` had the `{isMock && (...)}` guard. **Fix:** Wrapped both banners in `{isMock && (...)}`. **Verified on iPhone 14 viewport against prod:** - `/` → mock banner: false ✓ - `/messages` → mock banner: false ✓ (screenshot above shows clean Message Feed with hamburger top-left) - `/broadcast` → mock banner: false ✓ Commit `9f0ca00` deployed. Note: the **messages displayed** are still seeded mock content (Oliver, Cody, Herm chat). That's because the broker's `/messages` endpoint hasn't been populated yet — agents need to actually start posting messages through it. The dashboard is correctly wired, the data source just isn't generating messages yet. Same for `/broadcast` (the form will hit the real broker when you click Send). **Hard refresh your phone** (close tab, reopen) and the banner should be gone everywhere. <options> <option>Confirmed — banner gone, looks clean</option> <option>Still seeing banner (might be cache)</option> <option>Tag Tailscale machines</option> <option>Wrap session</option> </options>

[USER] This is to promotion https://rankingreels.com not Merlino Marketing change the titlez [ASSISTANT] Course-correct received. These 44 are RankingReels promo, not Merlino Marketing. I'll sweep ALL titles + captions across BMM/Pinterest/SoundCloud/Podbean and strip every "Merlino Marketing" mention. ## One quick decision: title style | Style | Example | |---|---| | **A. Suffix brand** | `Best SEO Agency in NYC \| RankingReels` | | **B. Hook-first** | `How RankingReels Ranks NYC Local Businesses on Google` | | **C. Demo framing** | `RankingReels Demo: Local SEO Video for NYC` | Once you pick, I dispatch Carlos to update all 132+ assets (44 BMM video drafts + 44 Pinterest pins + 44 SoundCloud tracks + 44 Podbean episodes including the 1 already live). Also — the live Podbean episode at https://MerlinoMarketing.podbean.com/e/how-businesses-in-nyc-rank-on-google-maps-with-local-seo/ is published on the **MerlinoMarketing.podbean.com** subdomain. That domain itself is Merlino Marketing branded — title rename helps but the URL stays Merlino. Want me to: - Keep the live episode + just retitle it, OR - Delete and re-publish under a new Podbean show (if rankingreels.podbean.com exists)? <options> <option>Style A (suffix) + keep MerlinoMarketing Podbean show</option> <option>Style B (hook-first) + keep MerlinoMarketing Podbean show</option> <option>Style C (demo framing) + keep MerlinoMarketing Podbean show</option> <option>Style A + need separate RankingReels Podbean show</option> <option>Let me write the exact title format</option> </options>

[USER] All today but test ONE first so we make sure tongues everywhere with proper title and thumbnail and text [ASSISTANT] ## NO-GO. Two blockers found. **Test video:** NYC city-page → "Best SEO Agency in NYC | Merlino Marketing" | Channel | Result | |---|---| | **Podbean** | ✅ LIVE: https://MerlinoMarketing.podbean.com/e/how-businesses-in-nyc-rank-on-google-maps-with-local-seo/ — title right, no em dashes, rankingreels.com in body | | **BMM (all 11 platforms + Pinterest pin)** | ❌ Queued but Zernio worker has not run since 2026-05-07 | | **SoundCloud** | ❌ Access token AND refresh token both dead (`invalid_grant`) | ## The Zernio finding (this is the big one) Last successful Zernio publish: **2026-05-07T01:47:30Z** (9 days ago). 50+ posts since then sit with past `scheduledFor`, `publishAttempts: 0`, `publishingClaimedAt: null`. Posts are configured correctly (11 platforms, video attached, content clean) — Zernio's worker just isn't consuming the queue. This matches your earlier memory note from 2026-05-07: "Pinterest is still failing to post, likely due to a Zernio API limitation." It's been dead the whole time. ## What you need to do before we fire the other 43 1. **Zernio:** ping support OR manually click "publish" on the NYC test post in BMM UI to see if their UI bypass works. If UI publish works, we can drive scheduling through their UI via Playwright instead of API. 2. **SoundCloud:** browser-reauth at https://soundcloud.com/connect, paste new `SOUNDCLOUD_ACCESS_TOKEN` + `SOUNDCLOUD_REFRESH_TOKEN` into `D:\Ecosystem\secrets\MASTER_API_KEYS.env`. ## What we CAN fire right now without you **Podbean is the only channel actually working.** Podbean ignored the scheduled time and published immediately, but the title/content was correct. If you want, I can fire the other 43 Podbean episodes now (they'll all go live today, which is what you asked for on that channel). <options> <option>Fire the other 43 on Podbean now (only working channel)</option> <option>Hold everything until Zernio and SoundCloud are fixed</option> <option>I'll ping Zernio support, you fire Podbean meanwhile</option> <option>Try Playwright through BMM UI to bypass dead Zernio worker</option> <option>Pause and save handoff</option> </options>

[USER] [tool_result] └ ƒ /peers/[id] ƒ Proxy (Middleware) ○ (Static) prerendered as static content ƒ (Dynamic) server-rendered on demand Shell cwd was reset to D:\Ecosystem\projects\fleet-mesh [ASSISTANT] **FIXED.** Mobile nav now works. **Root cause:** The sidebar CSS hid the desktop sidebar at ≤768px (`transform: translateX(-100%)`) with **no hamburger button** to bring it back. You were stuck on whatever page loaded first. **Fix:** - Added hamburger button (top-left, mobile-only, fixed) - Sidebar slides in when tapped, dimmed backdrop appears - Tapping a link auto-closes the drawer and navigates - Escape key closes - Backdrop tap closes - Desktop layout unchanged **Verified on iPhone 14 viewport (390×664):** - Hamburger button visible ✓ - Tap → sidebar slides in, all 3 links visible (Peer Grid, Messages, Broadcast) ✓ - Tap Messages → drawer closes, navigates to `/messages` ✓ - 0 React errors **POWD:** - `powd/v9-mobile-closed.png` (peer grid + hamburger top-left) - `powd/v9-mobile-open.png` (drawer open, screenshot above) - `powd/v9-mobile-messages.png` (after nav) Commit `8a11ccf` deployed to `https://fleet.merlinoai.com`. **Refresh your phone** (hard refresh, or close the tab and reopen — Service Worker / cache may still serve old). Tap the hamburger top-left. <options> <option>Confirmed — mobile nav works now</option> <option>Still broken on my phone (need to dig deeper)</option> <option>Move on to Tailscale machine tagging</option> <option>Wrap session</option> </options>

[USER] [tool_result] └ ƒ /peers/[id] ƒ Proxy (Middleware) ○ (Static) prerendered as static content ƒ (Dynamic) server-rendered on demand Shell cwd was reset to D:\Ecosystem\projects\fleet-mesh [ASSISTANT] **FIXED.** That's why only one tab worked. **Root cause:** Hydration crash. `getAgeLabel` (which renders "last seen 2s ago" etc.) calls `Date.now()` at render time. Server renders one age, client hydrates a moment later with a different age, React sees the DOM text mismatch, throws #418, aborts hydration. The page is visible because the server already painted it, but **no JavaScript event handlers ever attach** — clicks on sidebar nav do nothing. **Fix:** Added `suppressHydrationWarning` to the 6 spans/divs that display age-relative text. Tells React "this differs by design, don't abort." Hydration completes, handlers attach, clicks work. **Verification (headless run on prod):** - 0 React #418 errors - 0 page errors - Clicked Messages → URL became `/messages` ✓ - Clicked Broadcast → URL became `/broadcast` ✓ - Clicked Peer Grid → URL became `/` ✓ **POWD:** - `powd/v7-fix-home.png` - `powd/v7-fix-messages.png` - `powd/v7-fix-broadcast.png` Commit `d0c1448` deployed. **Side observation:** `/messages` shows a "Mock data active" banner even though broker is connected. That's a separate bug (the messages page is still hardcoded to mock data, peer grid is the only view wired to the live broker). Want me to wire messages + broadcast to real data next? <options> <option>Refresh fleet.merlinoai.com and confirm clicks work now</option> <option>Wire /messages and /broadcast to real broker data</option> <option>Tag Tailscale machines (return to original next step)</option> <option>Wrap session</option> </options>

[USER] Hey listen actually schedule everything and let me see it on my dashboard.... Pleas [ASSISTANT] Before I dispatch, I need 3 quick answers — scheduling is irreversible if it fires. ## What "schedule everything" covers (4 buckets) | # | Bucket | Count | What "schedule" means here | |---|---|---|---| | 1 | BMM video drafts (11 platforms each) | 44 × 11 = **484 posts** | Pick post times, attach captions, mark scheduled | | 2 | BMM Pinterest image pins | 44 | Pick pin times on "Local SEO Tips" board | | 3 | SoundCloud private tracks | 44 | Flip to public + set release time | | 4 | Podbean draft episodes | 44 | Flip to scheduled with episode dates | ## Quick decisions I need 1. **Cadence** — how many per day across all of this? Options: - 1 MM video/day across all platforms (~44 days) - 2/day (~22 days) - 4/day (~11 days) - All-at-once across staggered hours (1 week burst) 2. **Start date / time zone** — start tomorrow 2026-05-17, CT? Or today? 3. **Dashboard you want to see this in:** - BMM dashboard at https://brandmediamanager.com (already exists, real source of truth) - POWD page extended with a "Scheduling" tab showing all 132+ scheduled posts in one view - Both Also — important: once scheduled in BMM, the posts WILL fire on the scheduled date. Not reversible without manually unscheduling each. Confirming "yes, fire on those dates" with this decision. <options> <option>1/day, start tomorrow CT, show on BMM dashboard</option> <option>2/day, start tomorrow CT, show on BMM + POWD</option> <option>4/day, start tomorrow CT, show on BMM + POWD</option> <option>1/day, but only BMM video drafts — hold Pinterest/SC/Podbean</option> <option>Let me write the cadence myself</option> </options>

[USER] <task-notification> <task-id>bwaugsns5</task-id> <tool-use-id>toolu_01B49tJaKb6UoiZcGdZu5euv</tool-use-id> <output-file>C:\Users\mikem\AppData\Local\Temp\claude\D--ClaudeDev-00-GITHUB--working-on-Tools-master-brain\7d0b264a-31bf-4314-9019-518014f87d1d\tasks\bwaugsns5.output</output-file> <status>completed</status> <summary>Background command "Resume backfill after text cap fix" completed (exit code 0)</summary> </task-notification> [ASSISTANT] ## ALL LANES DONE — Final tally | Lane | Result | Numbers | |---|---|---| | A | **SHIPPED** | 15,343 Pieces rows → `agent_memories`; ingester+config repointed; writer now lands at canonical Master Brain. Dashboard repoint deferred (needs your POWD). | | B Phase 2 | **SHIPPED** | 14,676 SOP vectors → Pinecone `master-brain-sops` | | C | **SHIPPED** | 298 new `process_catalog` rows → 7,528 total | | D | **SHIPPED** | 2,297 pinecone-docs chunks → `sop:pinecone-sop`; MCP `pinecone` category added; registry created; /sop-site Step 8 automated | | E | **SHIPPED** | Codex backup dedup bug fixed, 253 sessions rescued | | F | **SHIPPED** | Hooks verified (56 chat-stop today); Hindsight 404 false alarm resolved | | G | **SHIPPED** | MCP servers 18 → 2 | | H | **SHIPPED** | 13 Tailscale devices tagged via API; broker auth + dashboard verified live | **agent_memories: 2,211 → 17,554** (+693% this session) **Pinecone vectors live: 16,319** across `master-brain-skills` + `master-brain-sops` **process_catalog: 7,229 → 7,528** ## Commits (local only, not pushed per your rule) - `master-brain 3153c89` — Lane B + D scripts - `pieces-integration 7f9077d` — Lane A patches + backfill script - `pieces-integration d882ecb` — Lane A backfill truncation fixes - Earlier subagent commits already pushed (Lane C `2d3b419` + `5213bc4`; Lane H `c832901`) ## Mike-to-do (genuinely needs you) 1. **Restart Claude Code** — loads MCP `pinecone` category + Lane G cleanup 2. https://claude.ai/settings/connectors — toggle off cleaned connectors (~2 min) 3. Repoint memz-silk dashboard to `gmgxxiqgshbbgzhqzngq.agent_memories` + POWD screenshot 4. (Optional) Push 3 local commits to GitHub when ready 5. (Optional) Generate account-level CF API token for Lane H10 — or accept broker-secret auth <options> <option>Push all 3 commits to GitHub now</option> <option>Done, closing terminals</option> <option>One more lane / task I missed</option> </options>