Ignacio Montenegro

Dear algorithm, I want to connect with people who are: - startup founders - solo founders - vibe coders - devs Drop a hi, let’s connect!

@MitcheIl Launch

@rodrigorojop @cjpedregal @meetgranola Thanks!!

@Ignaci0m_ @cjpedregal @meetgranola They usually allow the same task and data, but make certain a apps work better. An example: an API would be nice for automations, CLI nice for local agents through terminal, MCP for AI Apps like ChatGPT or Claude.

There are some tweets out there saying that Granola is trying to lock down access to your data. Tldr; we are actually trying to become more open, not closed. We’re launching a public API next week to complement our MCP. Read on for context. A couple months ago, we noticed that some folks had reversed engineered our local cache so they could access their meeting data. Our cache was not built for this (it can change at any point), so we launched our MCP to serve this need. The MCP gives full access to your notes and transcripts (all time for paid users, time restricted for free users). MCP usage has exploded since launch, so we felt good about it. A week ago, we updated how we store data in our cache and broke the workarounds. This is on us. Stupidly, we thought we had solved these use cases well enough with our MCP. We’ve now learned that while MCPs are great for connecting to tools like Claude or chatGPT, they don’t meet your needs for agents running locally or for data export / pipeline work. So we’re going to fix this for you ASAP. First, we’ll launch a public API next week to make it easier for you to pull your data. Second, we’ll figure out how to make Granola work better for agents running locally. Whether that’s expanding our MCP, launching a CLI, a local API, etc. The industry is moving quickly here, so we’d appreciate your suggestions. We want Granola data to be accessible and useful wherever you need it. Stay tuned.

@Surreal_Intel @AnthropicAI Could you give me some examples of the old open source infrastructure that AI Frontier uses?

@AnthropicAI Good move. The interesting point is that frontier AI does not float above the stack. It sits on open-source infrastructure that is old, shared, and often underfunded. Securing that layer is not charity, it is system maintenance.

The open source ecosystem underpins nearly every software system in the world. As AI grows more capable, open source security becomes increasingly important. We're donating to the Linux Foundation to continue to help secure the foundations AI runs on.

@archiexzzz @karpathy Sorry if this is too basic a question, but can I run this on the cloud, or do I need a GPU?

Introducing AutoVoiceEvals I've applied the @karpathy autoresearch loop to voice AI agents. It's open source. Your voice agent has a system prompt. That prompt determines how it handles every call - bookings, complaints, edge cases, background noises, long pauses, people trying to trick it. Most teams write it once, test manually, and hope for the best. autovoiceevals makes it a loop. One artifact (system prompt), one metric (adversarial eval score), keep what improves it, revert what doesn't. Run it overnight. Wake up to a better agent. > How it works: You describe your agent in a config file - what it does, its services, policies, and what it should never do. You don't write test cases. You don't define attack vectors. provider: vapi / smallest ai assistant: id: "your-agent-id" description: | Voice receptionist for a hair salon. Maria does coloring only. Jessica does cuts only. $25 cancellation fee under 24 hours notice. Cannot advise on skin conditions. Closed Sundays. From that description alone, Claude generates adversarial caller personas - each with an attack strategy, a voice profile (accents, background noise, mumblers, interrupters), a multi-turn caller script, and pass/fail evaluation criteria. The eval suite is generated once and held fixed for the entire run, like a validation set. > The loop: 1. Read the agent's current prompt from the platform 2. Generate adversarial eval suite from your description 3. Run baseline 4. Claude proposes ONE surgical change to the prompt 5. Push the modified prompt to the agent via API 6. Run all scenarios against the updated agent 7. Score improved? Keep. Same score but shorter prompt? Keep. Otherwise revert. 8. Go to 4. Run until Ctrl+C. The system sees its own experiment history. When a change fails, the next proposal knows what was tried and why it didn't work. We ran 20 experiments on a live Vapi dental scheduling agent. 0 human intervention. > Score: 0.728 → 0.969 (+33%) > CSAT: 45 → 84 > Pass rate: 25% → 100% > 9 kept, 10 discarded > Prompt: 1191 → 1139 chars (better AND shorter) You describe your agent. It figures out how to break it.

My new favorite word

@mariodian @karpathy @_kaitodev Could you share the link?

@karpathy @Ignaci0m_ @_kaitodev Thanks for the clarification! I forked it just minutes before you took it down because I wanted to add some minor UI tweaks. The fork gained a bit of traction now... Do you want me to take it down?

5 minutes ago, @karpathy just dropped karpathy/jobs! he scraped every job in the US economy (342 occupations from BLS), scored each one's AI exposure 0-10 using an LLM, and visualized it as a treemap. if your whole job happens on a screen you're cooked. average score across all jobs is 5.3/10. software devs: 8-9. roofers: 0-1. medical transcriptionists: 10/10 💀 karpathy.ai/jobs

@karpathy @_kaitodev Any chance you can upload it again or share it with me? I’d love to have access to it.

This was a saturday morning 2 hour vibe coded project inspired by a book I’m reading. I thought the code/data might be helpful to others to explore the BLS dataset visually, or color it in different ways or with different prompts or add their own visualizations. It’s been wildly misinterpreted (which I should have anticipated even despite the readme docs) so I took it down.

found this site terraink.app and i'm honestly obsessed with how these look. you can make these super clean maps with totally different vibes. i've been playing with the neon style for Berlin and then this classy gold and blue theme for Paris and New York. if you want a cool wallpaper or just like street grids you gotta try it. @DataScientistPy

was moving my memory from ChatGPT over to Claude and found some really creepy things in the data logs... it has me down as 'Flagged as low current risk with theoretical risk if influenced or radicalised' and says I should be 'considered part of Emerging Civilian Assets & Trends (ECAT) class.' WTF?? why is it profiling me like this? has anyone else found weird tags like this in their memory export??

The future is here. Individuals can now move faster than big pharma!

@shayaan_azeem @cursor_ai Is it in-person?

we’re hosting @cursor_ai freeform a 4 hour build night with no themes or tracks just come build something you actually care about and ship it by the end of the night reply or dm if you want in

@NotebookLM When are we getting the option to select specific sources or multiple ones instead of just "all" for the chat and artifact generation (audio overviews, video overviews, flashcards, and so on) in the iOS app? Also, when are you giving the share options so I can share with other people and change the visibility of the notebook? I'm getting a persistent error when trying to change the visibility from restricted to anyone with the link and also when trying to add new people to the notebook.

Finally Google is using its ecosystem advantage on the AI race!!

@tylerbruno05 @Lovable Add support for building iOS Apps

I just joined Lovable as their first Technical Evangelist. My job now: push the limits of what's possible with Lovable (and share it all here), embed with users, and bring what I learn from you back to the team to help shape what's next. That starts with office hours. Recurring, open to anyone. Share what you love, tell me what could be better, bring your feature requests, and wildest ideas. First slots open tomorrow and I'll drop the signup link here. What would you build if nothing technical stood in your way?

@Base44 @MS_BASE44 I tried DMing you but it gives me an error message

@Ignaci0m_ @MS_BASE44 If you're ready to test our take on agents, DM us your email address to be added to the list!

We’ve been working on something big. It’s our take on agents the base44 way (as always - batteries included) I’ve been using it non stop, and am so excited about this it’s literally hurting my sleep. Still in alpha, and looking for early feedback Comment below or repost if you want access, will also give credits to those who can deliver valuable feedback

@nestymee What do you use for the virtual try on?

>$10k/mo app idea personal stylists charge $200/hr. built one that lives in your phone for $15/month and knows every piece you own aesty.ai