Why is Astra so good at 3D modeling? Better RL envs vs dropping English CoT
willdepue · x · 2026-09-07
- Hypothesis (akarvonen): Astra's big jump in 3D modeling comes from strong no-CoT capabilities — reasoning in English tokens has long seemed like a bottleneck for spatial reasoning.
- Counterpoint (willdepue): "it's probably just better RL envs honestly, I'm pretty sure."
- He adds that Astra's perception also improved markedly — likely mostly scale, but likely some pretraining architecture/encoder wins too: "Astra can see really well and this surely helps."
Related event: Debate Over Astra's 3D Leap: No-CoT Ability or Scale?(3 posts)→
More from Models
- Games Are Making Astra Go Viral: Long-Horizon Coherence, Fast Computer Use, and Taste — manosaie · 2026-09-07
- Fable 5 vs 5.1: same prompt, one word apart in the ending — and sadder prose — repligate · 2026-09-07
- Viral demo claims GPT-6 Astra researched references to render Silo scene in Blender — Eric520CC · 2026-09-07
- Fable 5.1 is a big step up in editing ability — and more work for its user — TheZvi · 2026-09-07
- Astra chat blocked by unoverridable safeguard, forcing users to restart — nrehiew_ · 2026-09-07
- GPT saturates SRE reverse-engineering benchmark, sparking debate on end of software ownership — dosco · 2026-09-07