AI-Assisted Development: 28,000 Lines in 13 Days
Published June 2026
ZestyShop went from zero to a fully functional e-commerce marketplace in 13 days. Not with a large team — with one developer, AI pair programming, and a clear architectural plan.
The Raw Numbers
Daily Commit Cadence
Development followed a burst pattern — intense feature pushes separated by brief pauses. The largest spike (28 commits on June 16) was the initial v0 scaffold. Subsequent spikes show iterative feature building.
| Date | Commits | Theme |
|---|---|---|
| Jun 16 | 28 | v0 scaffold — auth, payments, emails, checkout |
| Jun 17 | 11 | ESLint, Sanity CMS, marketing pages, wishlists |
| Jun 19 | 10 | Theme toggle, account consolidation, shipping demo |
| Jun 20 | 23 | Multi-vendor marketplace, merchant dashboard, orders admin |
| Jun 22 | 20 | Aurora seeding, DB ops, merchant templates |
| Jun 28 | 13 | PGlite, content builder, category images, UI polish |
| Jun 29 | 6 | Architecture diagram, articles, inventory, dark mode fixes |
Assumption — burst pattern
The gaps between bursts (Jun 18, 21, 23-27) likely represent planning and debugging phases where fewer atomic commits were made, or days where work happened outside the repo. In AI-assisted workflows, the pattern is common — rapid generation followed by review and consolidation.
What Got Built
Looking at the commit subjects, every core e-commerce feature was delivered in sequence:
- Day 1-2 — Auth (Better Auth + Google OAuth), Stripe/Dodo payments, SES emails, checkout flow
- Day 3-4 — Admin CMS, product management, category system, content builder
- Day 5-6 — Multi-vendor marketplace, merchant dashboard, storefront pages
- Day 7-8 — Order management, shipping tracking, admin notifications
- Day 9-10 — Database operations (Aurora seeding, PGlite local dev), test suite
- Day 11-13 — UI polish (empty states, 404 page, animations, dark mode), documentation
Commit Type Breakdown
Features dominate at 48% — a hallmark of greenfield AI-assisted development where the model generates entire modules from a prompt. Fixes and refactors account for 20%, reflecting the iterative polish cycle.
Contributors — Human + AI
Three entities contributed code to the repository. The mix reveals the AI-assisted workflow:
- v0agent — Generated the initial scaffold: auth, payments, emails, checkout, pricing pages. 25 commits from the AI agent in the first hours.
- kit — Human developer. 11 commits covering ESLint setup, Sanity CMS configuration, and metric pages. The human handled tooling and infrastructure.
- CodeDoes — AI pair programming hybrid. 72 commits (65% of total). A developer using AI as a coding partner — generating features, fixing bugs, refactoring, and iterating at high speed.
Assumption — the AI pair programming ratio
The 72 commits attributed to CodeDoes represent a developer working with AI code generation tools. Each "commit" likely contains a mix of AI-generated boilerplate, human-written business logic, and AI-assisted debugging. The high feature-to-fix ratio (3.8:1) suggests AI handles the initial write; the human handles review and integration.
Why This Matters for the Hackathon
The H0: Hack the Zero Stack challenge is about shipping production-quality software fast. ZestyShop demonstrates that the Vercel + Aurora PostgreSQL stack, combined with AI-assisted development, turns a weekend project into a deployable marketplace. The 28,618 lines of code include:
- Full product catalog with search, filter, and sort
- Multi-step checkout with discount codes
- Email/password and Google OAuth authentication
- Stripe subscriptions and DodoPayments checkout
- Multi-vendor marketplace with merchant dashboards
- Admin CMS for products, orders, content, and theme
- 165 unit tests and Playwright end-to-end coverage
- Dual database setup: PGlite for local dev, Aurora PostgreSQL for production
The tools let you build fast. The architecture makes it scale. AI makes it possible in a weekend.