Coinbase post-trained a 9B model to beat Opus 4.5 on fraud review, reproducible on Modal
AAAzzam · x · 2026-10-14
A widely shared thread revealed that Coinbase post-trained a 9B-parameter model that beat Anthropic's Opus 4.5 at fraud review at much lower cost and latency; the author also showed how to reproduce it on Modal with minimal setup.
The poster built on this with a take: "owning your intelligence" is cultural in nature — the goal is driving the marginal effort of customizing models per use-case down to zero, so companies default to their own Opus rather than a generic frontier model.
More from coding & agent
- When building is 10x faster, knowing when to slow down matters 100x more — brandon_galang · 2026-10-14
- Matt Pocock: compacting the orchestrator session is safe because its task is simple — mattpocockuk · 2026-10-14
- Building games with AI? Start with a graybox playable before final art — chongdashu · 2026-10-14
- Gemini 4 Argon goes live on Vercel AI Gateway at 50% off as rauchg hails coding agents as AI's killer app — AccBalanced · 2026-10-14
- Dev shares slash-command combo: prototype with /ce-prototype, then build with /lfg — kieranklaassen · 2026-10-14
- Kevin Rose hails Claude 'Projects' beta as a 'beautiful orchestration of sub-agents' — joellewenstein · 2026-10-14