Meta-agent: self-improving agent harnesses from live traces
Iteratively improves agent harnesses from 67% to 87% on tau-bench using production traces.

Intent-based resolution tracking beats span-level observability for AI agents.
Teams shipping customer-facing AI agents in production
Langfuse · LangSmith · PostHog
Iteratively improves agent harnesses from 67% to 87% on tau-bench using production traces.
Buffer + AI video generation — promises learning, ships generic scheduling.
Finally sees the bot traffic GA4 filters out, with citation attribution.
Direct DOM access beats screenshot agents—milliseconds not seconds per action.
Production-ready CUDA profiling when NSight only works in development.
Camel Camel Camel meets Honey, but unproven against established deal aggregators.