3x depth means 81x params: xuanalogue and Greenblatt debate the causal role of depth scaling
xuanalogue · x · 2026-09-04
Ryan Greenblatt and xuanalogue debate the role of depth in architecture scaling: the claim under discussion is that scaling 'normally' yields 81x more (active) params per 3x depth increase. xuanalogue argues the causal direction is unclear — OpenAI could have scaled width instead, and recurrent depth may merely let depth grow proportionally with width rather than drive the gains. A fundamental question about whether depth scaling independently helps.
More from Research
- Aviation tracks every part's provenance — agents can reason over that graph — colinmcnamara · 2026-10-21
- SFT details: instruction/reasoning/code/math mix, synthetic + human traces, no epoching — TimDarcet · 2026-10-21
- Mid-training recipe disclosed: 32k→131k context extension, dropping logit distillation for tokens — TimDarcet · 2026-10-21
- Mathematician Daniel Litt concedes his AI bet, but says AI won't solve math — ziv_ravid · 2026-10-21
- AI drug discovery reality check: 'give me everything that binds,' then filter for toxicity — burny_tech · 2026-10-21
- PyTorch details KTIR and Dataflow Scheduler behind IBM's Spyre accelerator backend — PyTorch · 2026-10-21