Why 12 months behind may be the floor for AI fast-followers, not the gap
herbiebradley · x · 2026-09-24
A debate on whether intelligence compounds into a durable frontier lead. The original poster models intelligence as having compounding returns, predicting no equilibrium of multiple comparable providers. herbiebradley pushes back: purchasable third-party training data, distillation, and the leakiness of frontier know-how make it actively hard to stay 12 months behind — he views 12 months as the lower bound for the gap, with AI R&D automation shrinking it further.
Related event: Debate: Buyable Data Makes a 12-Month AI Lag Hard to Sustain(3 posts)→
More from AGI Musings
- Raising VC will feel like a sign of inefficiency as AI lets companies build core tech in-house — eherrerosj · 2026-09-24
- Gary Marcus posts receipts: same AGI promises keep slipping, Musk's 2025 pledge included — GaryMarcus · 2026-09-24
- Is part of the population simply unable to recognize AI writing, even daily users? — robleclerc · 2026-09-24
- When a reporter asked if models could take over the internet, Vishal Misra's rebuttal got cut — vishalmisra · 2026-09-24
- Commenter: academic denial of AI impact lasts 6-8 months, biology will fall next — IgorCarron · 2026-09-24
- The Great AI Debate lands on Diary of a CEO with opposing AI futures panel — romanyam · 2026-09-24