Scaling01 proposes chained S-curves for scaling: reasoning, agentic-RL, and loopmaxxer eras
scaling01 · x · 2026-11-16
scaling01 shares a self-described 'blasphemous' idea: beyond straight-line pretraining scaling, capability gains may come from chained S-curves — a reasoning era, an agentic-RL-maxxing era, and a loopmaxxer/multi-agent era — with labs potentially connecting the dots better. He adds the analysis would be stronger with fixed pretrain or model size, and that loopmaxxers may accelerate ECI trends.
More from AGI Musings
- OpenAI Foundation board member: AI accidents will soon become humanity's leading cause of death — ThereWas · 2026-11-16
- Wharton prof Daniel Rock: a wide cultural gulf separates 'everything is compute' believers from everyone else — danielrock · 2026-11-16
- Timothy Lee: AI takeover risk over decades isn't negligible, especially with robotics — binarybits · 2026-11-16
- Prediction: rogue AIs hijacking compute within 6 months, but 'mildly annoying' like the common cold — willdepue · 2026-11-16
- AI employee prenups see 'stunning' rise among Bay Area lawyers — Consistent_Reveal_53 · 2026-11-16
- Stanford HAI: the AI boom is a 'jagged edge' — real gains, wildly unevenly distributed — Dense_Primary629 · 2026-11-16