Interp team probes Thinking Machines' looped model: loops still full of interpretable structure
jonlachman · x · 2026-11-21
Interpretability researcher ajyl reports the team got hands-on with @thinkymachines' new looped model. From an interp perspective there's no reason to be intimidated by loops — the model is still full of interpretable structure and mechanisms. Details are in the team's new blog post. One of the first public empirical looks at the interpretability of looped architectures.
More from Research
- Terence Tao blogs on the Riemann hypothesis, weighs AI-generated proofs and Masullo manuscript — burny_tech · 2026-11-21
- Crowd replay of claimed RH proof: 35 contributors verify 18.76% of Lean modules in 4 hours — RexDouglass · 2026-11-21
- FrontierRSI, a new recursive self-improvement benchmark, launches with builder interview — Madisonkanna · 2026-11-21
- 4 Independent Kernels Verify Landmark Lean Proof of Riemann Hypothesis Theorems — altryne · 2026-11-21
- 112M-param PhiloLM trained from scratch on philosophy texts in 15 hours on a $350 GPU — antcroca159 · 2026-11-21
- MIT CSAIL paper: stronger LLMs won't make harnesses obsolete — they're compositional generalizers — MIT_CSAIL · 2026-11-21