Archives
All the articles I've archived.
-
Karpathy: the eval game has lapped the model game
By Nizam Alavi ·
Karpathy's latest thread argues frontier labs are now optimizing for benchmarks more than capability — and that the way out is harder, weirder evals nobody can game.
-
Sam Altman is publicly preparing investors for a 'compute cliff'
By Nizam Alavi ·
A subtle but unmistakable shift in Altman's messaging: from 'compute is abundant and getting cheaper' to 'compute is the binding constraint and will be for years.'
-
Simon Willison's quiet tool shipping is doing more for indie AI dev than any frontier release
By Nizam Alavi ·
While the labs argue about benchmarks, Simon is shipping small CLIs that make the actual day-to-day of working with LLMs unrecognizable from a year ago.
-
Dario Amodei's quiet pivot from 'safety-first' to 'safety-economics'
By Nizam Alavi ·
Read the last three Amodei essays back-to-back and the argument morphs: from 'we slow down because it's right' to 'we slow down because it's the only sustainable economic posture.' That's a different story for different audiences.