Rox trains an outbound reward model on seller edits: 5x faster, 30x cheaper than LLM judges, 1.75x reply rate
andrew_ilyas · x · 2026-10-13
Sales AI startup Rox shared how they built the feedback loop commercial agents never had: coding agents improved by writing code and getting feedback, so Rox trained an Outbound Reward Model on real cold emails plus the edits sellers make to AI drafts before sending.
- 5x faster and up to 30x cheaper than an LLM judge
- Picks the seller-edited draft 80% of the time
- Top-scored emails get 1.75x the average reply rate, vs 1.33x for a GPT judge and 1.23x for a Claude judge
The chart shows the reward model consistently outperforming both LLM judges as fewer emails are kept.
More from coding & agent
- Animating a cheetah via physics simulation reveals the limits of rigid-body skeletons — Kyrannio · 2026-10-13
- Running local AI agents filled a 1TB MacBook in three months — davidfromkansas · 2026-10-13
- 15 AI agents racing to notify me about the same email — omooretweets · 2026-10-13
- Every's guide to voice-driven agents: patch bugs and write by talking — every · 2026-10-13
- Delegating to your agent at 7am and going for a run: how do you count work hours now? — gethackteam · 2026-10-13
- Developer teaches Claude to paint in 3D, framing it as a pure-code motion-solving experiment — Kyrannio · 2026-10-13