KosmosCompactBench benchmarks compaction algorithms; Prime Intellect best open source
BenBlaiszik · x · 2026-11-14
Researchers released KosmosCompactBench to compare context-compaction algorithms from frontier labs against 5 open source harnesses. Results show truncation collapses in multi-turn settings, OpenAI native compaction beats Anthropic native on LongMemEval multi-turn (86% vs 33%), and Prime Intellect's open source compaction is the strongest open implementation, which the team adopted as their starting point for autoresearch.
More from coding & agent
- We don't need more chatbots: an agent harness lead's field notes — threepointone · 2026-11-14
- Nexus launches an AI context engine that turns deployment knowledge into agent-queryable graphs — rohanpaul_ai · 2026-11-14
- Developer now runs 100% of his agents locally on a Mac Mini, up from 25% earlier this year — josh_bickett · 2026-11-14
- Mediabunny ships 400 fixes from an LLM-assisted bug hunt across its 20M-download codebase — Vjeux · 2026-11-14
- Perplexity Teammates screenshot shows agents on Datadog duty, reading PRs overnight — testingcatalog · 2026-11-14
- Nexus launches as an agentic workforce for AI deployment teams — dr_cintas · 2026-11-14