With CliffCompaction, open-weight models beat closed ones in long-horizon sessions
Tim_Dettmers · x · 2026-09-23
Tim Dettmers observes that CliffCompaction creates a usefulness gap between open-weight models and closed models that hide their reasoning traces: for long-horizon sessions, open-weight models are far stronger, and some models simply "compact better" than others. This builds on his point that the tool works much better with full thinking traces — consistent with frontier labs, where OpenAI's math and ARC-AGI results largely stem from trace-aware autocompaction.
More from coding & agent
- Harvard, Andrew Ng and Karpathy Teach AI Engineering Free: A 9-Step Learning Path — mdancho84 · 2026-09-23
- Cookie banners vs agents: X user shows Europe's web friction makes AI agents look superhuman — jeff_weinstein · 2026-09-23
- New ComfyUI nodes let you mask any video area and insert characters — roychodraws · 2026-09-23
- Shopping agents shouldn't delete shared ingredients when you cancel one meal — Electrical-Yard-287 · 2026-09-23
- From 1 Month of Runway to YC: Founder Sold Computer-Use 'AI Employees' on Orgo — nick_linck · 2026-09-23
- New Claude Opus Spontaneously Uses Temp Symlinks to Fix Rust Image Pipeline — generativist · 2026-09-23