Show HN: Microcosm – a substrate AI coding agents read to orient before acting

Category: devtools

Tags: ai-workflow, developer-tools, project-intelligence

Score: 6.8/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 5)

Microcosm is an AI-native workflow substrate that generates a reproducible, inspectable map of a project's architecture, evidence, and claims without modifying source files or calling external models. It innovates by encoding evidence classes, source links, and scope limits into a fixed grammar, making the system's claims auditable and rerunnable, though its niche as a one-person research prototype limits broader utility.

Target audience: backend devs, devops, data engineers

Repository: https://wcook04.github.io/microcosm-substrate/ · Python · Apache-2.0 · 2 stars

View on Hacker News