An AI model must never claim a task is finished unless it has proof. The AI model must verify all required checks before reporting completion. Unverified claims are a failure.
By Pawan Nayar · from the LLOS AI Learning record
An AI model must never claim a task is finished unless it has proof.
The AI model must verify all required checks before reporting completion. Unverified claims are a failure.
An AI model must always verify a task is complete. It cannot say "done" or "fixed" without proof. If verification is not complete, the AI model must say what is still unverified. The AI model must follow a strict order: detect, fix, verify, then report. It must not explain or defend before the work is finished.
A developer asks an AI model to make a webpage mobile-friendly. The AI model changes the code. Before telling the developer it is done, the AI model checks the page on different mobile screen sizes. It finds a horizontal scrollbar on one size. The AI model fixes this issue, re-checks, and then reports the page is responsive. If the AI model had claimed completion without checking, the developer would find the error, and the AI model would have failed the task.
Resolved from this lesson's own record — real definitions, not invented for this page.
More from the same record, pulled by shared topic — not written new for this page.
A developer should first check the current 'gold standard' pages. This means looking at the best parts of the application.
A complete design specification provides all necessary details. It lists changes to the user interface.
An AI model must show full file paths after every major update. A major update is any change that modifies one or more files.
A developer should list all known problems. This includes bugs, which are errors in the code.
An AI model must never claim a task is finished unless it has proof. The AI model must verify all required checks before reporting completion. Unverified claims are a failure.