Timeline

Speaker 1: It's got a— Speaker 2: What are you running? Speaker 1: It's got a stealth mode on it, I think. I think they have stealth modes as well. But it does... it can support proxies too.

Speaker 1: It's got a stealth mode on it, I think. I think they have stealth modes as well. But it, it can support proxies too. Speaker 2: But I'm saying in the free GitHub, this Crawl4AI shit is going to scrape the internet not using my IP, 'cause you know...

[Chat Transcript] Rich Fong's Personal Meeting Room Participants: Garret Acott, Brian Hong, Andrew Ansley, Chris Labbate, Matteo, Michael Merlino, Rich Fong. You (11:44 AM): Uncle G can I get api access to Turbo? Also the GitHub you shared is not usable, is it an idea? Brian Hong (11:53 AM): https://github.com/paperclipai/paper Andrew Ansley (11:54 AM): qwen/qwen3-vl-32b-instruct https://huggingface.co/Qwen/Qwe-VL-32B-lnstruct Chris Labbate (12:01 PM): As of mid-May 2026, ZoomInfo Technologies (GTM) has a market capitalization of approximately $1.15 billion to $1.84 billion, reflecting a significant decline from its 2020 IPO valuation of over $13 billion. The stock has faced pressure due to lowered revenue guidance, weak growth, and intense market competition, with shares recently trading around $4-$6. Matteo: I'm not going to use or repeat that token. You should revoke/rotate it in Hugging Face now because it was pasted into chat. Treat it as exposed.

I was using Crawl4AI and it's on GitHub. I'll find it, but it doesn't use any proxies, just straight through the internet. And I think it's using either like, you mentioned Firecrawl...

Speaker 1: ...through the internet and I think it's using either like... you mentioned Firecall? Speaker 2: Straight through the internet. Speaker 1: So real quick, because I'm installing it and my guys, you know, they're trained, right? They're like, "Yo, we're not running this, confirm what proxies you want." Which means it's probably running...

[Command-Line Output] [honcho] post-tool-use captured: PostToolUse:Bash says: craw14ai-setup 2> (success) Bash(craw14ai-doctor 2>&1 | tail -20; where crwl) Installed and verified. Done - Installed craw14ai 0.8.6 via pip (Python 3.13) - Health check passed: live crawl of https://craw14ai.com succeeded in 1.79s - CLI available: crwl at D:\Program Files\Python313\Scripts\crwl.exe Heads up: - The /install command I tried to run is wired for a TAC-2 style project. I bypassed the template and installed craw14ai directly. Quick usage: import asyncio from craw14ai import AsyncWebCrawler async def main(): async with AsyncWebCrawler() as crawler: result = await crawler.arun(url="https://example.com") print(result.markdown) CLI: crwl https://example.com -o markdown Dep conflicts to know about: Pip flagged two pre-existing conflicts (not caused by craw14ai): - oci 2.166.0 wants cryptography<46 and pyOpenSSL<=25.1.0 - vibevoice 0.0.1 wants transformers==4.51.3 (have 4.45.2) Proxy reminder: craw14ai = scraping tool. Use proxies (D:\Ecosystem\secrets\) for any bulk/repeated crawling. Want me to scaffold a project folder (E:\Codex\projects\craw14ai-playground\) with a venv + example scripts + proxy config wired in?

Speaker 1: I'm sorry, somebody mentioned the crawler earlier. What proxies are you using with that, just out of curiosity? Anything? Whoever mentioned something? Somebody mentioned a scraper earlier. Speaker 2: It's called Crawl4AI, and it's on GitHub. I'll find it, but it doesn't use any proxies, it's just straight...

Speaker 1: Freaking crazy. It's, um... yeah, it's pretty cool. Speaker 2: That with Flowbox would be killer, right? Speaker 1: I mean, yeah, it's all right there. It's just a couple Lego blocks just waiting to attach each other.

Speaker 1: And then it'll kick out responses to that too. Um, it's just freaking crazy. It's, uh, it's pretty cool. Speaker 2: That with Flowbox would be killer, right? I mean, yeah, it's all right there. It's just a couple Lego blocks just waiting to attach each other. Speaker 1: Hey, so...

Screen: to o main car los select • Enter Conduct Caliber Digital internal build to view 7m 47s [honcho] user-prompt injected conclusions (cached) L UserPromptSubmit says: GO received. Handing to Carlos. carlos(Conduct Caliber Digital internal build) Write( iber-digital- idebar\app-sidebar . tsx ) L Bash(python -c import os...) Running... Write( iber-digital- ( auth ) \cal . tsx ) +44 tool uses (ctrl+0 to expand) (ctrl+b to run in background) * Scaffolding Caliber Digital internal dashboard... (8m Scaffold Caliber Digital internal dashboard 32s to 12.2k tokens • thought for 7s) bypass permissions on (shift+tab to cycle) • esc interrupt ctrl+t to hide tasks • to manage Say GO and I hand to Carlos to conduct. [honcho] response saved response (2275 chars) L Stop says: * Baked for 1m 44s x recap: Build

So we kick out the phone scripts and then we'll do the follow-up responses and then also the objections. How do you handle a potential—we'll look at the potential objections as well and then it'll kick out responses to that too. Um, it's just...

It's just like, I mean we kick out like on the phones, it's like, it's uh, so we kick out the phone scripts and then we'll do the follow-up responses and then also the objections and how do you handle a potential, we'll look at the potential objections as well.

Speaker 1: And the amount of money here is just nuts. Speaker 2: Yeah, the AI... Speaker 1: And because, you know, that, that kind of, that degree of information we're getting is just like, I mean, we kick out, like on the phones, it's like, it's a...

Speaker 1: ...deals for 50k plus. I mean, and we're not even working it hard. It's like, holy shit, man. The amount of money here is just nuts. Speaker 2: Yeah, totally. Speaker 1: And because of that kind of, that degree of information we're getting...

Speaker 1: ...one to two billion because of the staff... Speaker 2: Yeah, the market is—the market is huge though, and the opportunity in that space is just freaking amazing. I mean, we're already picking up, you know, deals worth 50k plus. I mean, and we're not even working it hard. It's like, holy shit, man.

Speaker 1: The pros, the one that I was thinking of. And at one point there were 13 billion, but now they're like one to two billion because of the crash. Speaker 2: Yeah, the market—the market is huge though, and the opportunity in that space is just freaking amazing. I mean, we're already picking up...

Speaker 1: Yeah. Speaker 2: All right, deals are being made. Speaker 1: There you go. Speaker 2: Not even an hour yet. Speaker 1: That's great. Speaker 3: Yeah, I posted their evaluation. ZoomInfo was the one that I was thinking of. And at one point they were 13 billion, but now they're like...

Speaker 1: We should definitely have a chat on. I'll send you a message on Facebook. What's the best way to reach you? Speaker 2: Yeah, Facebook's good. Facebook's good. Speaker 1: Yeah. Alright, deals are already being made. Speaker 2: There you go. Speaker 3: We haven't even been here an hour yet. Speaker 1: That's great. Speaker 2: Um, yeah. I posted their evaluation. Zoom...

[Command-Line Output] North Valley DONE: - Translation Bridge REST API live (Railway + translate.merlinoai.com) - 5 sales-page variants live (Agenforce, Schedule, Proactiv, AI SaaS, ShipFast) - Chooser hub live at tb.merlinoai.com - 4 FAL Nano Banana 2 - Cory + Global NEXT: - Verify tb.merlinoai.com with neutral Style A-E tabs - Infographics deployed across all 5 sites - Mike (founder) photos deployed across all 5 sites - Rules locked: NO EM DASHES, light theme default, no curl to Mike - tb.merlinoai.com SSL is good, share with Cory, pick a winner - Open items: email-capture wiring, DEVTB_API_KEY lockdown, optional FAL #3 regen SAVED TO: - Auto memory (5 new memories + MEMORY.md index updated) - Context bundle (.agents/context-bundles/bundle-2026-05-15-15h.md) - .continue-here.md (.planning/.continue-here.md) - WIP commit (worktree branch infographics-fat-rerun) To resume: Open a new Claude Code session in D:\Ecosystem\tools\Development-Translation-Bridge\ and run /resume .planning/.continue-here.md from the worktree. Quick links to share with Cory: - Hub: https://tb.merlinoai.com - Variants: https://tb.merlinoai.com/l through https://tb.merlinoai.com/5 - API: https://translate.merlinoai.com

Speaker 1: ...data, you could do the outreach, then we can go sell this as a service for businesses, for agencies, for roofers, for window people. I got a bunch of profit share JV deals. Let's definitely have a chat. I'll send you a message on Facebook. What's the best way to reach you? Speaker 2: Yeah, Facebook's good. Speaker 1: Facebook? Okay, I'll send a message.

[Command-Line Output] North Valley Cory Tasks: - bash: missing pages (15 cities) - bash: service slug renames + sub-service routes - Structural mirror: 105+ pages [honcho] response saved (2686 chars) L Stop says: Cooked for 19m 28s. Financing/warranty sub-pages, /contact-us, /support, no-indexed. Next action: confirm whether to close the remaining URL gaps (slug renames, /solar-panels-rights) across all 15 silos. [honcho] user-prompt injected conclusions L UserPromptSubmit says: Closing the rest. Dispatching 2 parallel agents. Running 2 agents... Phase 4A — Service slugs + financing/warranty L Read: lib/area-data.ts Phase 4B — Support + Rights + contact-us L Write: app/[city]/support/page.tsx Running Queen visual QA... (9m 59s, 22.9k tokens) Queen visual QA pass Phase 4A — Service slug renames + financing/warranty sub-services Phase 4B — Support + Homeowner Rights + contact-us alias +8 completed

[Command-Line Output] North Valley Cory v24.7.o v3.13.12 Claude Code Q. 1.142 Opus 4.7 (1M context) D:\ClaudeDev Claude Max https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct 240 skill descriptions dropped [doctor for details

Great responses, both on email and on phone. I can find the companies in the area. If we provide the data, you can do the outreach, and we can go sell this as a service for businesses, for agencies, for roofers, for window people. I got a bunch of profit share JV deals.

[Web Browser Content] Hugging Face: Qwen/Qwen3-VL-32B-Instruct. Model: Qwen3-VL-32B-Instruct. Description: Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date. Key Enhancements: Superior text understanding & generation, deeper visual perception & reasoning, extended context length, enhanced spatial and video dynamics comprehension, and stronger agent interaction capabilities. License: apache-2.0. Downloads last month: 1,446,432. Model size: 33B params. Tensor type: BF16.

[Command-Line Output] North Valley Cory Claude Code v24.7.o v3.13.12 Q. 1.142 Opus 4.7 (1M context) D:\ClaudeDev Claude Max hf_LWriYvxnRqWrK0bWovrjhuYNoATeEyjBUu bypass permissions on /effort

[Chat Transcript] Rich Fong's Personal Meeting Room Participants: Garret Acott Brian Hong Chat: [11:44 AM] You: Uncle G can I get api access to Turbo? Also the GitHub you shared is not usable its an idea? [11:53 AM] Brian Hong: https://github.com/paperclipai/paper [11:54 AM] Andrew Ansley: qwen/qwen3-vl-32b-instruct https://huggingface.co/Qwen/Qwe-VL-32B-lnstruct [12:01 PM] Chris Labbate: As of mid-May 2026, ZoomInfo Technologies (GTM) has a market capitalization of approximately $1.15 billion to $1.84 billion, reflecting a significant decline from its 2020 IPO valuation of over $13 billion. The stock has faced pressure due to lowered revenue guidance, weak growth, and intense market competition, with shares recently trading around $4-$6.

Speaker 1: Like for roofing, for lead gen, for anything? Or is it working? Speaker 2: 100%. I'm telling you the responses—so we're getting great responses both on email and on phone. Speaker 1: Oh man, I think we got a company we could form together, Garrett. I'll provide the data...

[Web Browser Content] Qwen / Qwen3-VL-32B-Instruct Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date. This generation delivers comprehensive upgrades across the board: superior text understanding & generation, deeper visual perception & reasoning, extended context length, enhanced spatial and video dynamics comprehension, and stronger agent interaction capabilities. Available in Dense and MOE architectures that scale from edge to cloud, with Instruct and reasoning-enhanced Thinking editions for flexible, on-demand deployment. Key Enhancements: - License: apache-2.0 - Downloads last month: 1,446,432 - Model size: 33B params - Tensor type: BF16

[Command-Line Output] North Valley Cory Source-of-truth: DDA Pattern What Happened NEW skill created: YMYL Blog prompt Page-type taxonomy (physical-gbp / child) - Added as Mode 5 City Swap Test - Added as audit pt 18 Information Gain Mandate (3 things AI can't say) - Added as pre-write gate Banned-words list (24 tells) - Added as hard filter @graph + sameAs to Wikidata - Added as new section Where: ymyl-blog-writer/ location-page-builder/ article-pipeline/ content-humanizer/ + at-content-sniffer/ schema-stack/ Skipped: ICP Generator, Brand Voice Creator, Subtopics, SEO Content Enhancer, Research Assistant, Blog Topic Generator, Technical SEO prompt. Deferred: DDA Content Process Guide format. Cleanup: 90 em dashes stripped across 7 files. Next: restart Claude Code, commit 6 changed + 1 new file in merlino-skills. [honcho] response saved Recap: Installed crawL4ai 0.8.6 globally with Playwright browsers. CLI 'crwl' is ready. Proxy reminder: Use proxies (D:\Ecosystem\secrets\) for any bulk/repeated crawling. Quick usage: import asyncio from crawL4ai import AsyncWebCrawler async def main(): async with AsyncWebCrawler() as crawler: result = await crawler.arun(url="https://example.com") print(result.markdown) asyncio.run(main()) CLI: crwl https://example.com -o markdown Dep conflicts: oci 2.166.0 and vibevoice 0.0.1 flagged.

Speaker 1: Show you what we got. Speaker 2: Can we do this B2B? Speaker 1: Yeah, we have 37,000 categories we track. Speaker 2: Do you think this will work B2B, B2C, like for roofing, for lead gen, for anything? Or is it... Speaker 1: 100%. I'm telling you the response... So we're getting...

Adjusted for your data for sure, but absolutely. Yeah, send me, send me a list and I'm happy to freaking run it and show, show you what we got because we can do this B2B. We have about 37,000 categories we track. So I've... I... do you think this will work B2B, B2C?

[Web Browser Content] Hugging Face - Access Tokens User Access Tokens + Create new token Access tokens authenticate your identity to the Hugging Face Hub and allow applications to perform actions based on token permissions. Active Tokens: - ava: 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 (WRITE) - claude: hf_LWriYvxnRqWrK0bWovrjhuYNoATeEyjBUu (WRITE) - herman: hf_...xLlP (WRITE) - machermes: hf_...pXjW (WRITE) - o- herc: hf_...EkBY (WRITE) - mattteo: hf_...Xfju (WRITE) Account: Michael merlino (mmelrino)

Speaker 1: ...independently or any other system. Speaker 2: 100 percent. Speaker 1: I'd love to do a test run here. How do we—can we get on separate calls? Speaker 2: Yeah, just send me something like that. I mean, I'd have to adjust it for your data for sure. But, uh, absolutely. Yeah, send me—send me a list and I'm happy to freaking run it and...

[Chat Transcript] Ava: Protect the 11:30 to 3:00 block because the calendar is packed straight through. [System/Agent Logs] - Ran today's meeting prep. - Sent Telegram intro + 3 standalone briefs for DGS Mastermind, Weekly Meet, and Merlino X Malone. - Saved the full brief at meeting-prep-2026-05-15.md, initialized processed-meetings-2026-05-15.json, and logged the run in 2026-05-15.md. - Cron creation is blocked: local OpenClaw gateway closed during verification. [Agent Activity] - Installing Craw14AI. - Attempting to install Qwen3-VL-32B-Instruct model. - Capacity check: VPS has 19GB free, model weights are 62.1GB. Download initiated to dedicated folder. - AgentBrain CLI installed and working locally; awaiting API base URL and API key for full activation.

Speaker 1: ...scoring system. Why I gave you all this is the idea is your system can convert this into opportunities better than like instantly or any other system. Speaker 2: 100%. Speaker 1: I'd love to do a test with you, Garrett. How do we—maybe we get on a separate call? Speaker 2: Yeah, just send me something like that. I mean, I'd have to...

Something like this where we have all these data points, number of children, credit score, interest, affinities, behaviors, do you like scuba, snow skiing, or a one through nine scoring system. If I gave you all this, the idea is your system can convert this into opportunities better than like Instagram.

[Web Browser Content] Hugging Face - Settings - Access Tokens User: Michael merlino (mmelrino) Access Tokens User Access Tokens + Create new token Access tokens authenticate your identity to the Hugging Face Hub and allow applications to perform actions based on token permissions. Do not share your Access Tokens with anyone. Existing Tokens: - hf_...pXjW (WRITE) - hf_...EkBY (WRITE) Create new Access Token: - Token type: Fine-grained, Read, Write - Token name: [Input Field] User permissions (mmelrino): - Repositories: Read/Write access to personal namespace repos - Inference: Make calls to Inference Providers/Endpoints, Manage Inference Endpoints Modal: Save your Access Token - Token value: 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 - Action: Copy, Done

...pixels on their website, but I don't need a pixel on a website. What I can do is I could just have a list based on segmented data. If I gave you something like this where we have all these data points, number of children, credit score, your interests, affinities, behaviors, do you like scuba, snow skiing, 139...

[Chat Transcript] Herman: Qwen3-VL-32B-Instruct install: blocked on VPS3 storage/runtime. I checked the model: https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct. Model is public / not gated. But the safetensor weights alone are 62.13 GiB. Current VPS3 space: 96G total, 66G used, 31G available. RAM: 7.8 GiB. GPU: none / nvidia-smi not found. So I did not start the download because it would fail halfway and waste bandwidth/disk. This box has ~31GB free and the model needs 62GB+ just for weights, realistically 70GB+ with tokenizer/cache/partial download overhead. Also this VPS cannot run the 32B VL model locally without a proper GPU setup. Plain blocker: Qwen3-VL-32B-Instruct full install does not fit on this VPS3 right now. Best next move: use a quantized GGUF/AWQ version if available, or attach/mount larger storage and run inference elsewhere. Ava: Protect the 11:30 to 3:00 block because the calendar is packed straight through. 9:00 AM ET run. Live mail/calendar used; ClickUp and Fireflies were snapshot-limited. Ran today's meeting prep. Sent Telegram intro + 3 standalone briefs for DGS Mastermind, Weekly Meet, and Merlino X Malone. Saved the full brief at meeting-prep-2026-05-15.md, initialized processed-meetings-2026-05-15.json, and logged the run in 2026-05-15.md. Cron creation is blocked right now. The local OpenClaw gateway closed during cron verification (Gateway TransportError, ws://127.0.0.1:18789), so the 3 pre-meeting reminders and 3 post-meeting action-item jobs should be treated as not armed. I saved the failure note at meeting-prep-2026-05-15-cron-error.md.

Consumer signals and intent data based on websites you visit, keywords you type in. If I can feed that to you, so for example, I'll just pull up a—so this is a law firm, we stuck a pixel on their website, but I don't need a pixel on a website because I can just have a list based on segmented data. If I gave you...

[Web Browser Content] Hugging Face - The AI community building the future. Create new Access Token Token type: Read Write Token name: [Input Field] User permissions (mmelrino): - Repositories: Read/Write access to contents of all repos under your personal namespace. - Webhooks: Make calls to Inference Providers and Endpoints. Save your Access Token: hf_HY0ZmtyNATsgDDZGCOVLdrbUcajjpQxLIP (Save your token value somewhere safe. You will not be able to see it again after you close this modal.)

[Chat Transcript] MacHermes: Craw14Al successfully crawled rankingreels.com and returned clean Markdown. Ava: https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct David Avocado Wolfe: We are Winning the Meme War. Panda: Command used: bash .local/bin/crwl https://rankingreels.com. It pulled: Hero copy, Nav links, Stats, Case studies, CTA section, Image URLs. So the install is working. Oscar: https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct Shane Francis: Karma Bazaar. [System/Terminal Logs]: - uv tool install 'huggingface-hub[cli]' - mkdir -p ~/Models/Qwen - du -sh ~/Models/Qwen/Qwen3-VL-32B-Instruct [Technical Analysis]: Model is ~67GB. You have 91GB free, so it fits but leaves only ~24GB. I'm going to install the HF CLI + download the model to a dedicated folder, not random cache. [Update]: Qwen3-VL-32B-Instruct install: blocked on VPS3 storage/runtime. The safetensor weights alone are 62.13 GiB. This box has ~31GB free and the model needs 62GB+ just for weights. Also this VPS cannot run the 32B VL model locally without a proper GPU setup.

I have for my product B2B data and so we track that. We'll actually use—I'm wondering if I can combine—I have the consumer signals and intent data based on websites you visit, keywords you type in. If I can feed that to you, so for example, I'll just pull up a—so this is a law firm we stuck a...

[Web Browser Content] Hugging Face - The AI community building the future. [Navigation] Search models, datasets, users... Models, Datasets, Spaces, Buckets, Docs [User Profile: Michael merlino (mmelrino)] Account, Authentication, Organizations, Billing, Repositories, Access Tokens, SSH and GPG Keys, Inference Providers, Webhooks, Papers, Notifications, Jobs [Access Tokens] User Access Tokens + Create new token Access tokens authenticate your identity to the Hugging Face Hub and allow applications to perform actions based on token permissions. Do not share your Access Tokens with anyone. [Token List] Name: macherl | Value: hf_...EkBY | Last Refreshed: less than a minute ago | Permissions: WRITE [Create New Token Modal] Token name: macherl Token type: Fine-grained Permissions: Read/Write access to repos, Inference Endpoints, and Inference Providers. [Token Creation Success] Save your Access Token: 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 Save your token value somewhere safe. You will not be able to see it again after you close this modal.

Speaker 1: ...things that we learned through those, that those avatars and that persona investigation is I think one of the biggest secret sauce. Speaker 2: So I have a API that I have for my product DB data and so we track that. We'll actually use... I'm wondering if I can combine... I have the...

Speaker 1: ...for this or whatever. But to reference to, subtly reference the things that we learned through that, those avatars and that persona investigation is, I think, the biggest secret sauce. Speaker 2: So I have a API that I...

Screen: Personal 22 Unread Unigram Search Matteo • typing Herc MacHermes • typing Herman 12:01 PM 12:00 PM 12:01 PM 12:01 PM Finished the interrupted Craw14Al test first, t... Ava V/ 11:59 AM https://huggingface.co/Qwen/Qwen3-VL-32... O Oscar V/ 11:59 AM https://huggingface.co/Qwen/Qwen3-VL-32... O David Avocado Wolfe We are Winning the Meme War. Panda 1 1:48 AM 2613 1 1:48 AM Herc typing Tod ay The main website resolves fine: bash https : / /agentbrain . sh/ # 200 0K So the CLI install is good, but the default API endpoint currently is not reachable/resolvable. Repo tests bash pnpm test Result: failed only because the repo has no test files: text No test files found Bottom line MTMSTORM joined the group via invite link Shane Francis Network 1 1:43 AM Karma Bazaar: Karma Bazaar "*CRESTOC 29 1 1:40

Screen: Personal 23 Unread Unigram Search Matteo Test passed. Herc typing MacHermes Test passed Herman typing Ava typing Oscar 1 1:59 AM I ran a real crawl from the install... 1 1:59 AM 1 1:59 AM Craw14Al successfully crawle... 1 1:59 AM 11:58 AM V/ 1 1:58 AM 1 1:48 AM 2613 1 1:48 AM Matteo bot • Initialized Craw14Al DB under /home/mike/.craw14ai • The setup checked apt browser dependencies; it reported 0 newly installed, 0 upgraded. Verified: • craw14ai-doctor passed • Python import passed • crwl CLI is available inside the venv Use it like this: cd /home/mike/.openclaw-matteo-jr/agents/matteo-jr/workspace source .venv-craw14ai/bin/activate crwl crawl https://agentbrain.sh -o markdown agentbrain.sh AgentBrain — Enterprise agent runtime The agent runtime that knows who can do what. Enterprise-grad

[Web Browser Content] Hugging Face - The AI community building the future. Michael merlino (mmelrino) Access Tokens User Access Tokens + Create new token Access tokens authenticate your identity to the Hugging Face Hub and allow applications to perform actions based on token permissions. Do not share your Access Tokens with anyone. Name: mattteo, Value: hf_...Xfju, Permissions: WRITE Name: herc, Value: hf_...WOZE, Permissions: WRITE Create new Access Token Token type: Fine-grained Token name: [Token name] Save your Access Token Save your token value somewhere safe. You will not be able to see it again after you close this modal. Token: hf_uvAJqTfOHFtHPgsCW0iqJTJfVfxJ1xEkBY

Speaker 1: Hey, you know, I just want to let you know they really share your passion for this, and we, you know, we noticed that, you know, you were, that you might be in the market for this or whatever. But to reference, to subtly reference that...