WhattaRAG is my on-device RAG app, built entirely in Swift, and one of the most useful AI engineering lessons I've picked up building it came from its eval suite. To make retrieval and prompt changes provable instead of felt, I built an eval suite with 242 test cases. Its biggest win wasn't confirming a fix, it was killing three plausible ones and proving the real bug is a missing computation, not a wording problem. Here's what the evals measure, what they ruled out, and why seven of eight gate clauses still fail.
Clients have always wanted it for yesterday, and developers have always cut corners to give it to them. That trap didn't go away with AI, it got 10x faster. If we can now write ten times more code, we can also ship ten times more bugs, and burn out ten times sooner. Here's how I think about spending that 10x: not on more features, but on the robustness, correctness, and user experience of the ones we ship.
How I cleared an App Store Guideline 4.2 (Minimum Functionality) rejection caused by an Apple Intelligence device requirement — without resubmitting a new build.
In a world where answers are everywhere, thinking is becoming a lost skill. Mindstorm: The Idea Generator App That Trains You to Think Better In a world where answers are everywhere, thinking is becoming a lost skill. You can ask ChatGPT anything, search on Google, or scroll endlessly—and you’ll always find answers. But here’s the