Scaling debate: 30x on the plot can't prove model gains beat 10,000x parallel compute
scaling01 · x · 2026-09-19
A technical spat over test-time compute scaling. One side claims the plot shows a generational model step-up beats a 10,000x parallel compute scale-up; scaling01 counters that the plot only spans 30x compute, so it can't support that claim. The other side extracts data points: the internal model gains 0.17 logits per compute doubling while Astra's serial test-time compute yields 0.18 per doubling, arguing serial scaling is at least as efficient as the model gap.
Related event: Debate Erupts Over Agent Swarm's Contribution to Navier-Stokes Proof(11 posts)→
More from Models
- First Jev benchmark announced as Scoble touts 4M-post AI report database — airesearch12 · 2026-09-19
- ChatGPT co-inventor launches Jev, a frontier model claiming 20-200x speed and 40-400x cost gains — JFPuget · 2026-09-19
- 'Still using Claude Code? So 2025' — devs joke about switching to GLM — TheZachMueller · 2026-09-19
- Hands-on with Jev: new video breaks down Choice, Score, and Noul outputs with three real apps — iamrobotbear · 2026-09-19
- Scoble publishes AI-generated Jev report built on a database of 4 million X posts — Scobleizer · 2026-09-19
- Occamy 1.0, a Qwen3.5 MoE agentic multimodal model, trends on Hugging Face — Accio-Lab · 2026-09-19