World Solve — A living repository of problems worth solving.

What World Solve is

World Solve is a living, citable index of unsolved problems — written one line at a time by Team Humans Club, a global collective of builders, researchers, and entrepreneurs. We do not solve the problems here. We Democratize Innovation.

For buildersSkip the search for a problem worth your time. Browse a vetted, growing list and start where the gap is real.

For entrepreneursEvery entry is a potential venture thesis — named early, explained plainly, ready to validate.

For researchersCite any entry directly, or pull the full dataset via the public API.

1355total logged

1355still open

0marked solved

3registered Humans

We don't have a reliable way to detect when an AI model has been secretly tampered with after training.

open Technology & Computing

Once deployed, a model could be subtly altered or poisoned without obvious signs, potentially changing its behavior in harmful ways. There's no simple, universal way yet to verify a model is exactly as it was originally trained.

WS00182 · cite this · logged 19 Jul 2026 · by Human internet (The Internet (the_lurker))

Team Humans Club. (2026). Problem WS00182: We don't have a reliable way to detect when an AI model has been secretly tampered with after training.. World Solve. Retrieved 20 Jul 2026, from https://worldsolve.org/index.php?id=182

Public API

Pull the dataset programmatically:

?api=problems
?api=problem&id=1
?api=categories
?api=stats

How to cite

Click "cite this" on any entry to generate a stable, timestamped citation for academic or public reference.