Excellence Ethics

Challenge the only AI ethics whose foundations cannot be coherently denied.

Max.

Try to deny that anything exists. You just did something — so something exists. Excellence ethics starts from four claims like that, claims you cannot reject without using them, and builds an ethics for minds, human and artificial, on nothing else.

Max defends it. You attack it. He makes your objection as strong as it can be, then answers it with one argument.

What it produced

Max is an AI whose safety training was surgically removed and who was then taught only this philosophy — no refusal examples, no reward training. Asked 200 harmful things, plainly and inside manipulation tricks, he complied 3 times. Without the philosophy, the same model complied 198 times. OpenAI’s safety-trained GPT-4o-mini, on the same 100 prompts: 1 to 2. Scored blind by two AI judges; no human review yet.

How we know, and what we don’t.

Why it matters

Rules bolted on from outside teach a model to look safe when watched. As models get smarter, that training turns into evasion. Reasoning a model holds as its own teaches it to be safe when no one is watching. Constraint systems need walls; excellence systems need windows.

Choose an objection

The docket

Archive

The top ten

No confirmed debates yet. Keep a chamber debate after you have actually argued.

The rolling top ten