Claude Code creator Boris Cherny: bet on general models, skip fine-tuning
rohanpaul_ai · x · 2026-09-28
Boris Cherny, creator of Claude Code at Anthropic, advised on Lenny's Podcast that the more general model will almost always outperform a more specific one. "Don't try to use tiny models. Don't try to fine-tune," he said, urging teams to bet on general models when they have the flexibility. He added that scaffolding typically improves performance by only 10-20%, and those gains often get wiped out by the next model release — so it's often better to just wait for the next one.
More from AGI Musings
- As fruit fly connectome demos go viral, a blogger probes the moral unease — genmon · 2026-09-28
- Author Kevin Bass calls the entire AI Safety industry an 'Orwellian, malevolent cult' — kevinnbass · 2026-09-28
- AI-picked catalyst dismissed by experts survived 1,000+ hours in acid — VraserX · 2026-09-28
- Andrew Yang warns AI will crush millions of jobs; Theo Jaffee pushes back with 200 years of automation history — herbiebradley · 2026-09-28
- Critic slams 'relevance-chasing' influencers reinventing themselves as AI prophets — PierceLilholt · 2026-09-28
- Beff Jezos slams AI safety establishment: create risks, then sell tokens as insurance — beffjezos · 2026-09-28