03 — Failure mode

Watch the model be confidently wrong.

Pick a prompt the model can't possibly know the right answer to (an out-of-date CPF rate, a fictional licence number, last week's PSF). Every token in the output is coloured by the model's confidence in that token. The reveal: high confidence and being right are different things.

Coming soon

Under construction

GPT-2 small in the browser; per-token logprob colouring. Audience-keyed prompt library that reliably produces fabricated facts in domain. Builds after the Tokenizer ships.