AI slop isn't a people problem, it's a process problem

Before CI/CD, deployment failures came down to individual responsibility. As the focus shifted from who pushed a bad change on Friday to why the pipeline didn't block it, responsibility for quality and stability moved from individuals to teams and systems.

The same shift is needed in AI-assisted coding. When something goes wrong after AI-generated code gets merged in, the first reaction is often to blame the developer for not being careful enough — but in reality, the hole was in the process that let that code pass through PR review and merge in the first place.

AI-assisted development is producing 2-5x more PRs. But the number of senior engineers available to review them isn't growing at the same rate, so having humans manually review every single line of AI-generated code doesn't scale.

The solution, as with CI/CD, is to codify standards and automate them. Rules should be written down — whether a new API endpoint needs rate limiting, whether adding an external dependency includes an explanation, whether DB queries follow the team's patterns, whether documentation was updated alongside code changes — and automatically checked on every PR.

Continue defines these checks as version-controlled markdown files inside the repo, letting an agent run the full set of checks on every PR. When the criteria are met, it passes silently; when they're not, it flags the specific issues with suggested fixes — enforcing the team's standards consistently without review fatigue.