← all posts

Why Visuals Beat Calls for Fixing AI Slop

Tayyab Javed ·2026-06-17 · 5 min read
#refactoring #process

Most agencies open a project with a discovery call. An hour of vague questions, a follow-up deck, and a quote you have to take on faith. When the actual problem is a codebase that's quietly falling apart, talking about it is the slowest way to make progress.

I do the opposite. Send me a repo and the first thing you get back is a visual — a dashboard that shows exactly where the rot is, scored and ranked, before money changes hands.

Calls hide the evidence

On a call, everyone nods. Nobody can point at the 1,400-line controller, the three competing date libraries, or the auth check that's copy-pasted into eleven routes. A visual makes all of it undeniable in about ten seconds.

If you can see the damage, you can price the fix. If you can only talk about it, you're guessing.

What the snapshot actually measures

I pull the repo into R2, run static analysis, and surface the things that actually predict pain: cyclomatic complexity hotspots, duplication, dependency sprawl, missing tests around the money paths, and the AI-generated patterns that look fine until they touch production.

Slop Index: 73 / 100  (high)
  - 18 functions over 80 lines
  - 3 date libraries (moment, dayjs, native)
  - 0% test coverage on /api/checkout
  - 11 duplicated auth checks

Then we talk — with proof on the table

Only after you've seen the visual do we have a conversation, and it's a short one because the hard part is already settled. You know what's broken, I know what it costs to fix, and the deposit covers the work — not a sales process.

Questions? Book a visual audit or ping me on LinkedIn.