Contact
FR EN DE
Our expertise
Apps & AI
Consulting & Support
Cloud & IT Architecture
About References Careers Blog
All articles
AI & Data

What AI didn’t do for me (2/4): the 20% that eat 80% of the time

80% of the product in 20% of the time. Then three round trips for a 13-pixel icon. Nobody ever talks about that second part.

What AI didn’t do for me (2/4): the 20% that eat 80% of the time

What I learned building my SSH client with AI: speed collapses exactly where you don’t expect it. Four real examples.

The icon that wasn’t sharp.

Three round trips on a 13-pixel icon. The problem wasn’t the drawing but the way it was scaled down to that size: the stroke ended up just under one pixel wide. Guaranteed blur, invisible on a screenshot but annoying every day in use.

The same flaw, found twice.

A button whose clickable area spilled across the whole width of the toolbar. To the eye, everything looked fine. The layout system was doing, by default, something other than what the code seemed to say. Fixed once, then found again elsewhere two days later.

The Cancel button that closed the application.

Cancelling a file transfer killed the whole program. The cancellation method looked correct, but it interrupted the work at a point where the library in use did not expect to be interrupted. The error didn’t bubble up cleanly: it took everything down with it.

A crash with a misleading message.

Opening a menu brought the application down, with an error message pointing at two elements that had nothing to do with the real problem. The cause: a way of declaring a menu that the tool accepts when you write it and rejects when you run it.

None of these four problems was solved faster by AI.

They were solved by reading, measuring and understanding what was really going on. That’s the part of the work AI doesn’t speed up.


“What AI didn’t do for me” series: 1 · 2 · 3 · 4

Author CAPCOD
Ready to start?

Let's talk about your project.

Tell us what you need, we reply within 48 hours.

Get in touch