Pathway's 150M-parameter BDH reasons in latent space, beyond Transformers
bigdata · x · 2026-09-27
Pathway CEO Zuzanna Stamirowska argues transformers aren't enough for true reasoning. Their BDH model (150M params) reasons in latent space, achieves ARC-AGI results at dramatically lower inference cost, learns continually via fast weights, and maintains state over long horizons. She contends long-horizon agents may require a fundamentally different architecture, targeting higher intelligence per watt via enterprise one-rack deployments, open weights and on-device AI.
More from Infra
- Harvard Puts Full ML Systems Curriculum CS249r Online for Free — techNmak · 2026-09-27
- SpaceX's vertical empire: from rockets to Colossus AI compute powering Grok — XFreeze · 2026-09-27
- Token-efficient reasoning model Swift-1.5-Qwen3.8-27B trends on Hugging Face — ukisai · 2026-09-27
- Hacker uses NVIDIA DGX Station to run local models for CAD, invites SF folks to join — hudzah · 2026-09-27
- Yacine goes 'absurdly bullish' on TensorTorrent in rare endorsement — yacineMTB · 2026-09-27
- DeepSeek V5 leak: 2T parameters, reportedly trained fully on Huawei Ascend chips — teortaxesTex · 2026-09-27