Why RiskLens CI Almost Broke Me

The Beginning Looked Simple When I started building RiskLens CI , the idea actually made sense to me. Trigger on a GitLab merge request Analyze code changes Return risk level, issues, and recommendations Conceptually… I understood it. But the system itself? Completely different story. It was running. The endpoint responded. And I thought… okay this should be working. It wasn’t. This Felt More Advanced Than Anything I’ve Built I’m going to be real here — this project felt advanced for me. Not because I couldn’t understand the idea… But because: The system behavior was unfamiliar The feedback loop wasn’t obvious And small mistakes didn’t break things — they just made them worse That part messed with me. The “Almost Working” Stage Is the Worst Everything looked like it was firing correctly: Webhook → ✅ Backend → ✅ AI processing → ✅ But the results? Slightly off Missing detail Not as “smart” as I expected That’s when frustration really started setting in. Because nothing was clearly broken