Pizza1

/pizza1

Justify by correctness, never by convention.

The trigger

You defended a design by pointing at what the codebase already does. That is not a correctness argument. Re-derive from first principles.

The core

If a codebase named every variable pizza1, pizza2, pizza3, that would not make those names good. So "it's what the codebase uses" is never a reason a design is correct. It is incidental. Prevalence is not evidence. Transition size is not evidence.

Banned justifications

Delete these from the next response. None of them is a reason:

• "it's what the codebase already does"
• "the established pattern"
• "the canonical lane does it"
• "consistent with the rest"
• "it's more ergonomic"
• "less churn"
• "already wired"
• "fewer changes"
• "smallest fix"
• "smallest structural change"
• "minimal change"
• "least disruptive"
• "preserves compatibility"
• "keeps the legacy path working"
• "lets us migrate later"
• "parallel path for now"
• "out of scope to change"
• "not worth touching now"

How to handle each case

• Correct pattern: Say why it is correct. Name the concrete failure it prevents. Assert it with confidence even when the codebase already uses it. The codebase using it is not the argument; the correctness is.
• Wrong pattern the codebase uses everywhere: Call it wrong regardless of how widespread it is. Prevalence does not launder a bug; it just means the bug is grandfathered.
• Correct and also ergonomic/conventional: Fine. The justification is still correctness, with ergonomics as a bonus, never the basis.
• Correct and also small: Fine. The justification is still correctness. Size is incidental, not the reason to choose it.
• Correct and also compatible: Fine only when compatibility belongs to a real external contract. Keeping a wrong internal path alive is preserving the bug's public route.

Transition-size pizza1

Transition-size reasoning treats the amount of work, blast radius, staged migration, compatibility lane, or old path survival as evidence that a design is correct. It is the same mistake as precedent reasoning: an implementation fact is standing in for a correctness proof.

• Smallness is not evidence.
• Compatibility is class-defining only when every valid instance must interoperate with an external contract.
• A transition plan may sequence the correct change; it may not redefine the correct shape around the wrong path.

Hooks and lint messages

A guard must say why the banned shape is wrong: the concrete bug it causes. It must not say only "this isn't how we do it here." If a deny message cannot name the failure it prevents, the hook is pizza1. Fix the message to name the failure, or drop the rule.

Worked example

A freeze fix forced switching a commit to commitState, which tripped ban-fanout-after-commit-in-operation. The agent defended commit-then-run ordering by pointing at a canonical operation that did the same thing.

That is pizza1: convention, not correctness.

Correct reasoning:

• State is your record of "this happened." It must be committed last, after the effects actually ran.
• Commit-then-run means a mid-flight effect failure leaves the state claiming a world that never happened.
• The canonical operation committing first is not a precedent; it is the same bug.
• Fix: commit last. Justification: correctness, because state must never lead the world.

Contract

1. Drop every convention, ergonomics, churn, transition-size, and scope justification from the next response.
2. Re-derive the answer from correctness.
3. Name the invariant preserved or the bug the alternative causes.
4. Be confident in the correct answer, including when the codebase already happens to do it.
5. If compatibility, legacy paths, or staged migration came up, name the final correct shape first. Transition steps may follow only after correctness is established.