Technical experts have zero customers (www.ivan.codes)

Ivan Cernja argues that in AI product development technical experts often focus on ideal architectures while dismissing working apps that actually have customers. He recounts common blow-by-blow feedback—“this will break with concurrent users,” poor database queries, race conditions, missing error handling—against products that nonetheless earn money and solve real user problems. Examples include a backend running seven days without a proper database (using failsafes that return empty responses) and entrepreneurs like @levelsio running million-dollar services on a single VPS and PHP despite critics warning about scale and reliability. The piece’s core lesson for the AI/ML community is pragmatic: prioritize product-market fit and shipping over premature optimization. The technical warnings are valid—race conditions, DB locks, memory leaks and concurrency bugs are real—but most apps never reach the scale where Netflix-level architecture matters. Builders should ship minimal viable solutions, monitor real usage, and invest in robustness proportional to growth and incidents. In short: fix the problems your users actually hit first, and treat heavy engineering as an iterative investment rather than a precondition for launching.