Agnost AI Launches Log-Fine-tuned Model: +22.9% Success, -94.5% Cost
ycombinator · x · 2026-08-21
Y Combinator-backed Agnost AI launched its first model, agnost--0.1, focusing on fine-tuning using production agent logs. The method demonstrated significant improvements in real-world customer tests:
- Task Success: +22.9%
- Latency: -90.2%
- Cost: -94.5%
The approach repurposes agent traces—typically used only for debugging—into training data to outperform general frontier models. The startup offers fine-tuning services to optimize models for specific production environments.
More from coding & agent
- Muse Spark 1.2 Generates Web Pages and Games Directly from Images or Video — AIatMeta · 2026-08-21
- Miles v0.1 integrates Mooncake backend, accelerating remote data fetch by 14x — BanghuaZ · 2026-08-21
- How to set up your own Claude Marketplace — cspenn · 2026-08-21
- AI-built app turns a year of Google Timeline into video, hits 1.3k GitHub stars — basedjensen · 2026-08-21
- Stripe & Netlify Live Demo: AI App from Prompt to Prod — thisiskp_ · 2026-08-21
- LangSmith Launches Preview Builds: Spin Up Temporary Deployments to Test Agents from PRs — LangChain · 2026-08-21