A directory of AI software, organised by what you are actually trying to do: categories, use cases, and the integrations a tool has to have to be any use to you. The taxonomy is the product. Anyone can list AI tools; the work is in making thousands of them navigable by someone who knows their problem but not the vocabulary.
An AI directory has a problem the others do not: it rots faster than it grows. Tools are abandoned constantly, and a directory full of dead links is worse than no directory. So the site checks the tools it lists and publishes what it finds: a running study of which AI products have already gone. The rot became the content.
The same instinct produced the research desk: mapping the tools in the directory against real job functions and public employment data, to answer which jobs AI tooling is actually aimed at. That is the kind of page that earns links on its own, which is what a directory in a crowded category needs.
Python and Flask, SQLite to build and Postgres to run, Tailwind on the front. The same stack as the other directories, and for the same reason: the interesting decisions were in the taxonomy and the automation, not the framework. Sold, and still running under its new owner.