Laminar releases flow-1, a model for catching silent agent failures
What happened
On October 20, 2026, Hacker News' front page reported that Laminar released flow-1, a reinforcement-learning-trained model that detects silent errors in agent traces. On Laminar's own benchmark, its detection quality matches GPT-6-sol at 23 times lower cost. Inside Signals, flow-1 searches calls and tool results, returns findings in a user-defined JSON format, and clusters similar events to build failure evaluation sets. The reports cover the release, the benchmark results and the Signals features.
Written by AI from the coverage · updated 3 hours ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageLaunch HN: Laminar (YC S24) – We trained a model to catch silent agent failures
Laminar 发布 flow-1，利用 RL 训练模型检测智能体轨迹中的隐性错误；其自建基准称检测质量与 GPT-6-sol 相当，成本低 23x。模型在 Signals 中检索调用与工具结果，按用户定义的 JSON 格式返回发现，并将相似事件聚类以构建失败评测集。
Heat over time
Heat now 9·Comparable peak 10（Oct 20 01:00）·Comparable change over 24 hours –
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.