The application of knowledge is hard work. Reading a book or listening to a podcast feels like progress. The effort required is minimal, and that’s why it’s so seductive. You can finish a chapter or an episode and tell yourself you’ve done something worthwhile. But applying

I shared a controversial take the other day at an event and I decided to write it down in a longer format: I’m afraid AI won't give us a "compressed 21st century". The "compressed 21st century" comes from Dario's "Machine of Loving Grace" and if you haven’t read it, you probably

Check out the new episode with Matt Klein! We dug into bitdrift and Envoy! youtu.be/iJlZ0tFzb84?si…

Prediction: software development will closely resemble the evolution of self-driving cars. - Considered for many years intractable, now run of the mill for a large subset of rides (robotaxi), all mobility within sight. - The key metric to watch is disengagement rate (i.e.: how

Some people today are discouraging others from learning programming on the grounds AI will automate it. This advice will be seen as some of the worst career advice ever given. I disagree with the Turing Award and Nobel prize winner who wrote, “It is far more likely that the

Honestly for as much hate as they get, OpenAI still has the best sense for what makes for a good ml product. Another killer feature on top of deep research, 4o image generation is actually such a nice add — big props

Dwarkesh’s episodes with Sarah Paine are some of the best educational and entertaining content out there. Massive cudos for bringing on such phenomenal guests Dwarkesh Patel

We didn’t win the @xAI hackathon this weekend - but we built something we’re really proud of for the Grokipedia track. It started from a personal moment: my mom shared a viral video claiming Donald Trump was buying pastries from a Persian store, and it wasn’t real - but she had

Incredibly refreshing to have people like Chris Lattner working on the lower parts of the inference stack, despite the seemingly homogeneous reliance on NVIDIA GPUs and CUDA. One day, we may see an end-to-end SOTA LLM trained and served on an AMD GPU cluster in the coming years,

I’m curious what the optimal AI + human in the loop coding environment is. There seems to be a gradient between coding from scratch on Vim to just running Claude Code and letting it vibe its way to completion. Do most people just use Cursor/Windsurf?

“A compiler did the work, but I vouch for the result.” With increasingly better LLMs, the above statement seems likely to become equivalent over time.

Constellation (Constellation Space) predicts satellite network failures before they happen and automatically reroutes traffic in seconds—no human intervention, no data loss. Congrats on the launch @kamranmajid11, Raaid Kabir, Laith, omeed! ycombinator.com/launches/P9f-c…

Fundamentally, what really is the difference between an RLM and S={context folding, Codex, Claude Code, Terminus, agents, etc.}? This is the last and most important RLM post I'll make for a while to finally answer all the "this is trivially obvious" from HackerNews, Reddit, X,

fwiw, I’ve been enjoying a mix of tmux + neovim + opencode. Feels like the right blend of operating out of the terminal and still having the ability to code/review with the agents.

Very much my experience tbh

Codex > CC

Introducing Fort, a wearable that automatically tracks strength training. Strength training is one of the best things you can do for your health and longevity. It deserves better tools.

Anu Bonthalapati Y Combinator Don’t forget constellation!

History in the making