Sonya Huang

Had a blast judging the @OpenAI Codex hackathon w @gdb @thsottiaux @lennysan & lobster king @steipete 🦞 It’s crazy what you can build in a single hackathon w coding models. Tonight’s finalists...

thought provoking read

harker forever ✌️

On today's Training Data, @zachlloydtweets breaks down what goes into building a great harness @warpdotdev: prompting; tools; knowing when to summarize, truncate, or spin up sub-agents; evals and o...

awesome work. the harness & the experiential data flywheel is the moat.

good list. peep @hwchase17 and how early he was. and oh no claude is officially better at my job than me 😢

The source of truth for software 1.0 is in the source code. The source of truth for long-horizon agents lives in the online traces themselves. @hwchase17 of @LangChain explains why: "With single L...

this question is central to the “will the foundation models eat the whole application layer” / “are all agents coding agents” debates.

sales led growth -> product led growth -> agent led growth

saddle up! let’s go!! 🐎

Alien-looking chip layouts on TPU... The equivalent to move 37 for chip design. Loved this episode.

As the digital realm becomes "solved" the moat moves to the physical world (humans, robots, networks, operations, etc).

yes i am a bene gesserit thanks for calling it out @barrald https://t.co/B7SsPD3VFp

It’s time to ride the long-horizon agent exponential. Today, your agents can work reliably for ~30m. But they’ll be able to perform a day’s worth of work soon – and a century’s worth of work event...

"So easy even a VC can do it" 😑 When @hwchase17 asked me to film a "dumb dumb and dumber" teaching VCs to vibe code agents, I was offended... and intrigued :P Awesome product. Congrats on the lau...

This week on Training Data: robots! I really admire @Physical_int for their open publishing spirit. @hausman_k and @jtspringenberg joined me and @Alfred_Lin to chat about pi*0.6, learning from ex...

Ravi is a ray of sunshine -- going to miss your huge smile and pearls of wisdom. But now's the time to get on the field. So damn excited for you and proud of you @GuptaRK22

Today on Training Data: Sanjit Biswas, founder & CEO of @Samsara (NYSE:IOT) and former @sequoia backed founder of @meraki Sanjit shares the ups & downs of running neural nets on constraine...