https://t.co/r3z1DhicuN

Cheetah was an early prototype of Composer-1. Composer-1 is significantly more intelligent while running at similar speeds!

The question I asked every day while working on Cursor 2.0: "does this change the way I build?" It does. Cursor 2.0 is a force multiplier for speed, insight, and reliability. Its built for a multi-agent world and powered by our own frontier model.

Why use LLM-as-a-judge when you can get the same performance for 15–500x cheaper? Our new research with @RakutenGroup on PII detection finds that SAE probes: - transfer from synthetic to real data better than normal probes - match GPT-5 Mini performance at 1/15 the cost (1/6)

You can find the talk on YouTube below. If this kind of thinking and focus on building great tools interest you, we’re hiring engineers across data, product engineering, and AI. Please reach out — scaling@thrivecap.com https://t.co/qNQgITeIlH

I've found it valuable to adopt good design ideas from "ordinary software" and apply them to AI systems. In the talk, I cover some of these principles that drive Puck's architecture, including the way we organize prompts and tasks, and our internal framework and logging tools.

Over the last year, by tightly integrating across our data stack, LLM infra, and our own workflows, we've been able to bring Puck's capabilities to a handful of surfaces across the team at meaningful scale. We inferenced around 10B tokens in October across 1000s of tasks.

Earlier this month, I gave a talk about some of the technology we’re building at Thrive & the philosophy behind it. One focus for us has been internal tools that augment our team's work, including Puck, an in-house research system that runs in chat & ambiently in the background.

C macros, Rust, and Python are all attempts at productivity improvements over programming in C, with vastly different tradeoffs. Productivity is not simply letting a human do the same work faster, and thinking about it differently often yields the most transformative new tools.

Instead of prompts to generate art, discover what human's have already created. Find art to match a feeling, a poem, or a dream… Explore with your imagination at https://t.co/wVN9y2YaRP

Spent three years looking for a team in Biodefense to invest in. But never found one. So we built it ourselves. Valthos builds next-generation biodefense. As AI and biotech rapidly advance, we're approaching near-universal access to tools with the potential to cure humanity or

My job is to be the guy in the cartoon, but super fast

From the shelf of @gucaslelfond

A lot of you here, for some reason I don't quite understand, have cared about my work for a long time — far before I thought it became worth listening to. I am reminded today how special that is. It has led to innumerable friendships and opportunities I am lucky to enjoy, which I

(!) with our team starting to close larger and more complex deals, @AmbrookAg is on the hunt for a founding sales lead. to date, we've grown to thousands of customers entirely without a formal sales team: everyone just learned how to do sales (including eng), and we built those

current rotation

I wrote some reflections on building AI products @_hex_tech for the last couple of years my biggest takeaway? internalize the bitter lesson and pivot your roadmap accordingly

🧵Sat down with @VentureBeat about the engineering approach behind Notion 3.0. Here's what I think not enough companies are willing to do, and why @NotionHQ never looked back

two years ago, i took a gap year to work at a startup in lake tahoe. as i was getting ready to return to school, josh’s tweet about the thrive fellowship caught my attention. candidly, i thought there was no chance i would be considered amongst the mass of talented T10

It's been really fun working with @jakeottiger. who started as part of our original Fellowship class. New applications are open — read more of his thoughts here.