Bindu Reddy says the industry still lacks a way to train 20T models and scale post-training RL
bindureddy · x · 2026-07-22
Bindu Reddy says he wants a Fable 6 or 7-class model, but scaling post-training is the real bottleneck
Bindu Reddy says Fable 5 is good, but he wants to see a much more capable Fable 6 or 7-class model. His bigger point is that the industry still has not solved how to train a 20T-parameter model and run millions of post-training RL loops at scale.
He argues that Meta or Google could probably do it, but suggests they may be distracted by making money from GPU sales.
More from AGI Musings
- AlphaFold Learned From 50 Years of Experiments, First-Principles Biology Still Needs Wet Labs — vishalmisra · 2026-10-26
- NYT roundtable asks five experts what the rules for AI companies should actually be — ShakeelHashim · 2026-10-26
- Anti-AI content is algorithm gold and very profitable, quips commentator — annetgriffin · 2026-10-26
- Silicon Valley breakdown: five Stanford-area giants worth ~$18T — FinanceYF5 · 2026-10-26
- Anthropic says Claude leads 26% of its AI R&D — but self-improvement still has limits — Exponential View (Azeem Azhar) · 2026-10-26
- Humanoids shattered forecasts: half-marathon record in 2026, 100m faster than Bolt — ben_j_todd · 2026-10-26