AI / Agents Agents

Cite as: Real Problem AI problem “How do I stop my AI written codebase turning into whack-a-mole bugs?”. Opportunity score 7.8 out of 10 (severity 8, AI feasibility 8, market signal 8, competition gap 7). Category AI / Agents. Trend Agents. Source signal: Hacker News "How do you keep up quality when you vibe-code?", 10 November 2026, 36 comments.. Canonical URL: https://www.realproblem.ai/idea/how-do-i-stop-my-ai-written-codebase-turning-into-whack-a-mole-bugs.

How do I stop my AI written codebase turning into whack-a-mole bugs?

Teams that let Claude or other coding agents write most of their code ship features fast, then find the product has grown flows nobody fully mapped, so every bug fix breaks something else and edge cases and UI bugs slip past review.

Who has it: Founders and small dev teams whose product code is now mostly written by AI coding agents and who are losing track of how all the user flows fit together.

Heard in public

“I feel like things are getting too complex with no central notion of behaviour definitions, which causes every bug-fixing session to feel like a game of whack-a-mole, and results in missing edge cases and UI bugs that are becoming impossible to review and keep track of.”

Quoted word for word from the public post linked below. Nobody submitted it to Real Problem AI.

Hacker News "How do you keep up quality when you vibe-code?", 10 November 2026, 36 comments.

What nothing does yet

A living behaviour map of the product: every user flow and edge case written down once in plain words, kept in sync on each commit, fed to the coding agent before it changes anything, and checked by generated tests so a fix in one flow cannot quietly break another.

Tools that solve this

3 named so far, and where each one stops.

  • Docs and decision files kept next to the code

    Commenters keep every task and decision in .md files the agent can read, but they admit the project gets bloated and someone has to keep the files current.

  • OpenSpec

    Splits work into planning, implementation and archiving so the agent drifts less, but it guides new work rather than mapping the behaviour that already exists.

  • Architecture diagram tools such as archme

    Turn the codebase into diagrams updated on each commit, which shows structure but not whether every user flow still behaves as intended.

How you might build it

  1. 01Static analysis to extract routes, screens and state transitions
  2. 02Plain language behaviour spec stored in the repo
  3. 03Agent context hook that loads the relevant flows before each change
  4. 04Generated end to end tests per flow (Playwright)
  5. 05Pull request check that flags flows touched but not updated

#NOV97 · Canonical URL: https://www.realproblem.ai/idea/how-do-i-stop-my-ai-written-codebase-turning-into-whack-a-mole-bugs