Reverse-engineering Generalist's Gen1.5: one continuous-time transformer unifying sensing, reasoning and action
KyeGomezB · x · 2026-08-20
Kye Gomez reverse-engineers a plausible architecture for Gen1.5 from public demos: a decoder-only transformer over continuous time with multi-rate sensor processing (vision, proprioception, force, tactile, language), a time-causal harmonic reasoning layer, and a flow-matching action expert generating 100 Hz trajectories plus a lightweight reflex loop. The key claim: sensing, reasoning and acting run asynchronously inside one model rather than via separate System 1/System 2 policies. He stresses this is a reconstruction, not a claim about the closed-source original, and argues such a capable model should be open-sourced.
Related event: Open-Gen: Open-Source Reimplementation of Generalist's GEN 1.5(3 posts)→
More from Embodied
- LLM-Controlled Robots Have <10% Chance of Reliability by 2028 — a_karvonen · 2026-08-20
- Using RealSense 3D camera for virtual green screen with depth control — chrismatthieu · 2026-08-20
- GEN-1.5 success driven by repetitive motion data and UMI collection — DrJimFan · 2026-08-20
- NVIDIA Tutorial: Post-Train Cosmos 3 Edge for On-Device Robot Control on Jetson Thor — MonaJalal_ · 2026-08-20
- Open-Gen Attempts to Replicate GEN 1.5 Embodied Model Architecture — KyeGomezB · 2026-08-20
- Three seconds to learn: Gen1.5's few-shot skill learning may hinge on one attention masking rule — KyeGomezB · 2026-08-20