One JSON is worth more than the whole graph.
If you mostly use chat-style AI and keep wondering which new tools are actually worth following, this is the mistake to avoid with Lum1104 / Understand-Anything [C001]: reading it as a graph viewer. That is the surface story. The more important read is that it is not selling visualization. It is selling a knowledge asset you can commit and keep with the code. [C002]
Why does that matter? Because if you only look at the pretty layer, you spend your attention on the wrong thing. A picture helps the person looking at it right now. A saved file can still help the next person who touches the repo later. The hidden cost is not just wasted attention. It is missing the step that makes understanding reusable.
The proof is in the docs-described behavior. /understand writes .understand-anything/knowledge-graph.json instead of stopping at a one-time visual. The README also describes committing that graph so coworkers can skip rerunning the full process, plus /understand --auto-update so the graph can stay current after new changes. That is a different category of value from "nice code map."
When you're judging a new AI tool, don't start with feature count. Start with this: does it change your next decision? Here, the signal is the reusable file, not the screenshot. If you know someone still judging tools by the demo, share this with them. Docs-only boundary: this is based on README-described behavior, not a hands-on runtime test.