Laminar launches flow-1 to catch silent agent failures
First merged summary.
Laminar says open-source flow-1 matches GPT-6-sol on its benchmark at 23 times lower cost.
Y Combinator S24 startup Laminar announced flow-1, a small open-source model post-trained with reinforcement learning to find implicit errors in agent traces, such as wrong clicks or citations when calls still succeed. The company says that on its benchmark flow-1 matches GPT-6-sol while costing 23 times less, so every production trace can be analyzed. flow-1 runs in a harness called Signals that treats spans like files, returns user-defined JSON findings, and clusters recurring events for eval datasets. Laminar’s observability stack is Apache 2.0 and self-hostable; revenue comes from managed cloud.
- Training used thousands of synthetic investigations across software, legal and support workflows.
- Signals can answer general prompts, not only raw error detection.
- Laminar says it does not yet trace voice SDKs such as OpenAI realtime.
- The company also cites 20x trace compression and a custom SQL engine.
What each outlet adds
beyond the shared facts- Hacker NewsReports Laminar’s flow-1 launch, benchmark cost claim, and Signals harness.
- Products
- SignalsRespan GatewayLiteLLM
- Organizations
- Y Combinator
Coverage
oldest first · one row per article · UTC- · 6d ago
Hacker News · securityFirst report
Laminar releases flow-1 to catch silent AI agent failures
Laminar says the RL-trained model matches GPT-6-sol detection quality at 23 times lower cost on its benchmark.
- · 5d ago
Respan launches open-source Rust AI gateway built from plugins
The pre-alpha gateway, under MIT or Apache-2.0, lets teams add providers and features without changing a small core.