Karpathy's notes after the Dwarkesh interview
AGI as a decade of agents, animals vs ghosts, RL as a straw, cognitive core, and why he does not want 1,000 lines of unsupervised code.
The interview went everywhere. This is Karpathy cleaning up the parts he botched by talking too fast: timelines, Sutton, RL, the cognitive core, nanochat, and a working relationship with models that is not 'go away for 20 minutes.' Bookmark-heavy on purpose.
Related reads
Grok Bots that design, simulate, and 3D-print
@ProfBuehlerMIT
MIT's Markus Buehler ran a team of Grok Bots from photos to a physics lab to a Bambu printer. The whole scientific loop, in one night.
Aug 22, 2026
24/7 app maintenance as a software factory
@dabit3
Nader Dabit on high-ROI agent workflows: four you can stand up in a minute, and forty more once the language is clear.
Aug 19, 2026
How loop engineering actually works
@adxtyahq
Mental model, production loop, and why the hardest part of AI is no longer the model. Video plus a write-up.
Jul 2, 2026
Autoresearch as a community of agents
@karpathy
Don't emulate one PhD student. Emulate a research community. Git is almost, but not quite, the right abstraction.
Mar 8, 2026
