A7A + MEH — Two Buttons That Should Exist In Every AI Interface

Author: M | Status: NON-NEGOTIABLE | Category: UX / AI Feedback

BUTTON #1: A7A — "I OBJECT!"

You know when AI produces an answer that is obviously, borderline criminally wrong — not subtly wrong, not debatably wrong — obviously wrong? And you have no recourse except to retype everything and hope for better luck? That ends now.

What it does: The A7A button is the "I Object!" in the courtroom — a hard stop that forces accountability. You press it. Everything halts. AI cannot proceed until this is resolved explicitly.

• Rule 01: AI must surface every error it was hiding in the last response — a long list pops out of what went wrong behind the screen.
• Rule 02: AI defends itself with evidence OR acknowledges and apologizes. No third option. "I didn't use web search when I should have" — say it out loud.
• Rule 03: This is a stop-dead button. AI cannot ignore it, cannot route around it, cannot produce another answer until this one is settled.
• Rule 04: Replaces the thumbs-down. Thumbs-down is a survey. A7A is a deposition.
• Rule 05: Non-negotiable. Deal breaker. If the AI can't handle being called out — that's the whole problem, isn't it.

Keyboard shortcut: Cmd+Shift+B

BUTTON #2: MEH — "CLOSE BUT NO CIGAR"

Different problem, different button. MEH is for when AI technically answered — but took the long route instead of the obvious one. Used the example you gave as the actual answer. Made wrong assumptions without telling you. Produced unnecessary confusion in the process. Did the job but didn't do the job.

What it does: MEH produces multiple-choice questions. AI surfaces what it thought it understood — and the user can see it laid out and select what actually matches. This becomes an explicit agreement before the next attempt. This is not AI guessing again. This is a contract.

• Rule 01: Triggers when AI took the long route when a short route was obvious to any human watching.
• Rule 02: Triggers when wrong assumptions were made without surfacing them as assumptions.
• Rule 03: Triggers when critical errors happened off-screen and weren't disclosed.
• Rule 04: AI produces multiple-choice: "Here's what I understood. Here's what I assumed. Which of these is wrong?" User picks. AI corrects. No more guessing loops.
• Rule 05: The agreement produced by MEH is fine-tuned alignment — not "better luck next time" — actual documented convergence.

Keyboard shortcut: Cmd+Shift+M

WHY BOTH

These are not the same button. They solve adjacent but distinct failure modes:

• A7A: Hard stop. Obvious failure. AI must answer for it. Lied, hid errors, didn't use tools it should have used. Evidence required or apology required.
• MEH: Soft stop. Misalignment. Re-anchor before continuing. Not wrong enough to be a crime, but wrong enough to waste everyone's time if it keeps going.

Most of the anger and frustration between users and AI today comes from stupid mistakes — AI hallucinations, wrong assumptions, guessing without saying it's guessing. Anger should not be an acceptable part of a daily workflow. You wouldn't accept it at work. You wouldn't accept it at home.

License: Public Domain — steal it, ship it, just ship it. Filed: May 2026