AI Pulse · Under the Radar

A new open-source tool called Understand-Anything has passed 80,000 GitHub stars by taking on a problem most software teams know well, getting a useful understanding of a large, unfamiliar codebase.

It turns a codebase into a queryable knowledge graph. Under the hood, it uses tree-sitter to parse the structure of the code, then adds an LLM layer that can help explain relationships and answer questions in plain language. It plugs into tools developers already use, including Claude Code, Cursor, Copilot and Gemini CLI. The project has 80,598 stars, 6,800 forks, and is gaining roughly 489 stars a day. With 110 open issues and 180 open pull requests, the activity looks like real developer engagement rather than a star-farm. You can see the project on its GitHub repository.

Why should a business owner care? Software knowledge is often trapped in the heads of a few experienced people. That creates slow onboarding, risky handovers, and expensive delays whenever a team needs to change an old system. Tools that can map a codebase and answer questions could make it easier for smaller teams to maintain systems they did not build.

The timing is interesting too. Graphify, another well-funded entrant pursuing a similar idea, is trending in the same week. That does not guarantee either tool will become standard, but it suggests code understanding is becoming a serious AI category. This is the kind of thing we build into an AI command centre, connecting business knowledge to the people who need answers quickly.