We built a Weights and Biases for autoresearch

Name: We built a Weights and Biases for autoresearch
Availability: InStock
Author: WecoAI

by WecoAI·Mar 20, 2026·2 points·0 comments

Visit Project View on HN

AI Analysis

●●SolidShip ItNiche Gem

Purpose-built tracing for agentic research loops when LangSmith feels too generic.

Strengths

•Targets multi-step agent loops specifically rather than generic LLM completion tracing.
•Run comparison tools help tune hyperparameters across autonomous research iterations.
•Shareable dashboards facilitate collaboration on experimental agent behaviors.

Weaknesses

•Launching via Twitter thread creates friction compared to direct product demos.
•LangSmith and W&B Weave already cover most agent tracing use cases.

Post Description

Hey, we're the team at Weco. We've been working on autoresearch tooling and kept running into the same problem — there's no good way to monitor what's happening across steps, compare runs, or share results with collaborators. So we built an observability layer for it — think W&B but purpose-built for autoresearch workflows: step-by-step monitoring, performance tracking, and shareable dashboards. Would love to hear how others are approaching this.