Judgment Labs launches publicly: monitoring and eval layer for long-horizon agents
testingcatalog · x · 2026-10-28
Judgment, an agent monitoring and improvement platform by Judgment Labs, is now publicly available. Users can trigger an investigation from Slack when an agent misbehaves; swarms surface similar failure cases, show affected use cases, and narrow down root causes. The platform proposes fixes tested against real production examples and tracks recurring behaviors to catch drift and regressions. It uses a data layer optimized for indexing large agent traces, cutting ingest costs while improving search latency and quality. Customers include Modal, Fluidstack, Listen Labs, Monaco, and Strala, with CTOs praising its eval and observability depth.
Related event: Judgment Launches Publicly as Monitoring Layer for Long-Running AI Agents(2 posts)→
More from coding & agent
- MASS paper: one model plays the whole agent team to bootstrap recursive self-improvement — idanbeck · 2026-10-28
- Dev builds AI-native macOS dock where an agent classifies activity and notifications in real time — trq212 · 2026-10-28
- AI-native macOS dock author: OS internals are painful, project not ready to share — trq212 · 2026-10-28
- MASS explainer: one model as orchestrator, workers and judge gains 1.2-1.6x score per token — idanbeck · 2026-10-28
- NousResearch revamps Hermes Agent plans: cloud agent included, daily/weekly limits — Teknium · 2026-10-28
- LangChain open-sources its design agent and deep-dives into how it was built — LangChain · 2026-10-28