# Decision Quality Is the Competitive Variable in AI-Augmented Builds ## Claim When AI handles code generation, the quality of a builder's decisions about problem, user, and scope determines outcomes more than time spent building. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - Holds when the AI tooling is capable enough to handle execution (code generation, UI scaffolding, debugging) without deep domain expertise from the builder. - Breaks when the technical problem is genuinely novel or complex enough that AI cannot generate working output without expert human guidance — frontier research, highly specialized systems work. - Time-boxed contexts (hackathons) make the effect visible fastest, but the claim isn't limited to them. ## Evidence Hong Darwin built a working AI-powered financial-health prototype for migrant workers in 48 hours with no prior coding background, using Lovable. Her own retrospective locates the actual constraint in decision quality, not code output: a single user conversation redirected her entire product concept, and a five-section PRD's clarity (or fog) propagated directly into prompt and output clarity. > [!quote] > "The biggest challenge was never the code during this buildathon. It was more about making good decisions under time pressure." ## Links - [[3. Resources/Literature Notes/How to Use 48 Hours Before the Hackathon]] — source literature note - [[3. Resources/Permanent Notes/Implementation Is Cheap So Taste Becomes the Bottleneck]] — because extends; same relocation of scarcity from generation to judgment, independently arrived at in a beginner no-code context rather than a professional-tooling one - [[3. Resources/Permanent Notes/The Bottleneck Moved From Code Generation to Review and Judgment]] — because extends; names the same shift from the reviewer's side rather than the founder's decision side - [[3. Resources/Permanent Notes/Creative Vision Sets the Ceiling on AI-Assisted Output Quality]] — because contrasts; that note's ceiling is aesthetic vision, this one's is strategic/product judgment — sibling claims about what AI leaves scarce - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — applies-in: Joins notes on how AI shifts value in building software. ## Tests / falsifiers - Compare hackathon outcomes for teams matched on coding output speed but differing in documented pre-build user research and PRD clarity. - If teams with the least pre-build decision work still win on demo/judging quality at comparable rates, the claim doesn't hold in this population. ## Maintenance - Last tended: 2026-08-16 - Next review: 2026-11-16 ## Source - [[3. Resources/Readwise/Articles/How to use 48 hours before the hackathon]] — Hong Darwin's 8-step pre-buildathon framework for SheBuilds/CBU Season 3 participants.