We Build Faster Than We Decide

Short summary

AI made building software faster, but deciding what to build remains the bottleneck. The author argues product teams should adopt a 'Software Scientist' discipline: form testable hypotheses, ship experimental code to gather data, track decisions and learnings in lab notebooks, and prioritize speed-to-learning over speed-to-code. This requires a mindset shift from treating shipped code as the primary output to treating user feedback data as the primary outcome.

• •AI shifted the bottleneck from execution speed to decision quality—the question is no longer 'can we build fast?' but 'do we understand the problem?'
• •Product engineering requires hypothesis-driven development with experimental discipline, not just faster delivery pipelines
• •Teams need persistent documentation ('lab notebooks') to track hypotheses, evidence, and learnings alongside code

Generated with AI, which can make mistakes.