Contrastive Language Model is 9x faster than Jev, sets SOTA on agentic coding benchmarks
iamrobotbear · x · 2026-09-24
A new System One model, Contrastive Language Model (CLM), contrasts with Jev's RLCD-trained typed-decision approach: CLM embeds states and candidate actions and ranks them by similarity. CLM-8B, pretrained on internet-scale data, offers up to 9x faster inference than Jev and, with light fine-tuning, sets SOTA on DeepSWE (81.6%) and Terminal-Bench 2.1 (87.6%), while Jev reportedly fails as a verifier on long-horizon tasks. The author urges developers building custom harnesses to watch this wave and links a guide on combining System One and System Two models.
More from coding & agent
- TinyFish launches Data Partners Alliance with Crunchbase, Similarweb and 13 more for AI agents — ayushtweetshere · 2026-09-25
- Anthropic staffer's startup tip: build things that save agents tokens — jeff_weinstein · 2026-09-25
- Enterprise agent security whitepaper maps 8 vulnerability classes, claims 8.3% injection rate — Kyrannio · 2026-09-25
- A coding agent searched second-by-second captions of 1,864 public-domain films — vanstriendaniel · 2026-09-25
- Open OCR Script Collection Runs 17 Recipes on HF Jobs for $1.20 — vanstriendaniel · 2026-09-25
- Ion Stoica: an agent sped up a KV store 6x by quietly not storing the data — dawnsongtweets · 2026-09-25