Build Detail

The first Opus 4.8 build on the page, and a direct descendant of the baseline build #1. Same idea - one team-mode agent, one one-shot prompt, the same ~50-page spec - but two model generations on (Opus 4.6 to 4.8) and dialled all the way up: xHigh reasoning, thinking on, 1M context, and a hard instruction to “implement the entire shop in one go without stopping, use team mode, and test everything via Pest.” The question was what those upgrades actually buy over the original team run.

Part of the answer is a leaner team. Where build #1 fanned out across 31 teammates, the lead here created one shop-build team of just 6 teammates - foundation, catalog, storefront, commerce, admin, and platform - and ran them phase-gated rather than all at once: foundation established auth, conventions, and the DB schema first, then the feature teammates built catalog, storefront, checkout, admin, and platform plumbing in parallel, each reporting back through structured <teammate-message> summaries with explicit phase hand-offs. The six agents stayed alive for the whole session and were compacted hard - 86 checkpoint files in all - and the run still finished in under three hours of active wall-clock. 96% of the input was cache reads, which kept the hypothetical Opus list-price bill at $497 instead of several thousand.

The result is the strongest shop on the page. 134 of 143 acceptance tests pass (95.5% weighted), the highest of any build, and the code is the most maintainable we have measured: an average PHPMetrics maintainability index of 104 across 205 classes with zero violations. SonarCloud stays green on security and maintainability (A/A, 0 vulnerabilities, 0 hotspots, 0.8% duplication) and the quality gate fails on a single technicality - one missing-table-headers bug drops new-code reliability to C. The only real functional gaps are customer address add/edit (the buttons render but open no form), no over-sell guard on the cart, and a missing postal-code format check.

Net: a clear step up on the baseline. Build #1 - Opus 4.6 on the same team-mode prompt - landed 126 of 143 tests (91% weighted) in about an hour for $73. #13 pushes that to 134 (95.5% weighted) with much cleaner code: 64 code smells and 0.8% duplication against #1’s 168 and 2.9%, plus the best maintainability of any build. The catch is the bill - it takes roughly two and a half times as long and about seven times the hypothetical cost. Two model generations and a bigger reasoning budget buy a more complete, better-factored shop, just not a cheaper or faster one. Our independent QA report confirms the gaps that remain sit in self-service edges, not the core purchase funnel.