Draft Slot

The original version was two repos, both dated early 2019. The API was Laravel 6 with Lighthouse GraphQL and Passport for OAuth. The frontend was React 16 with Apollo Client for data and Redux for local UI state, bundled with a hand-rolled Webpack config.

I picked every piece for a reason: GraphQL was new, Apollo was new, and picking them meant I'd learn them. That was the entire point of a thesis project.

• Laravel 6 + Lighthouse GraphQL on the backend
• React 16 + Apollo + Redux on the frontend
• Laravel Passport for OAuth2
• Hand-rolled Webpack config — no create-react-app
• Minimal tests, no deployment pipeline, no users

It delivered the thesis. It did not deliver a product. I submitted the work, graduated, and the repos went quiet for almost two years.

When I came back to it, I didn't port the old code. I rebuilt the whole thing in a single Laravel monolith with Blade templates, Alpine.js for interactivity, and Tailwind for styling. The GraphQL layer was gone. The SPA was gone.

The choice looked unfashionable on paper. In practice, it was the only one that made sense. I had no team, no budget, no investors, and a full-time job. Every hour I spent tuning a bundler or fighting hydration was an hour I wasn't spending on features my users could actually see.

• Laravel + Blade for server-rendered HTML, no hydration tax
• Alpine.js progressively enhances specific components
• Tailwind keeps CSS out of the critical path
• One repo, one deploy, one mental model

This was the version that actually shipped. Real users, a real database, a real draft night — the baseline the rest of this story is built on top of. The tradeoff was real too: I gave up some of the polish a proper SPA offers. What I got back was velocity. Shipping something on a Sunday night became normal again.

The Reset shipped, but it didn't do everything I wanted. I took the app that was already live and rebuilt its frontend and backend through an agentic engineering loop — requirements first, agent build, validation before merge. The product never restarted; the codebase did. Today the Laravel codebase is 100% built with Claude.

The goal of the rebuild was concrete: take the single-sport, server-rendered baseline and build the draft out into a real multi-sport, real-time product that scales. I got there by combining three threads, run through the same loop over and over.

• A full test suite written with TDD — the spec came first, the agent wrote to it, validation gated the merge
• Infrastructure that scales up and down per season — Laravel Vapor with three scaling profiles (offseason, pre-draft, draft week)
• An interface of modular components built in atomic fashion (Atomic Design), each one run through requirements to validated build

The numbers are the shape of a codebase someone has actually lived with through an agentic loop: 1,800+ commits, 347 Pest tests, and not a single missed draft night since the rebuild. The product predates this era. The code that runs it doesn't.

What the rebuild produced is a real product running across the NFL, NBA, MLB, and NHL. The work splits into three jobs that all have to hold at once: keep draft night live, treat the unglamorous plumbing seriously, and keep the bill sane in the months nobody is here.

Live drafts, in real time. On draft night every user watches the same board, and every pick has to reach every connected browser in seconds.

• Laravel Reverb for WebSocket broadcasting, with public channels per draft (draft.nfl.2026)
• Laravel Echo on the client subscribes; each pick is a broadcast event
• ESPN API polling as the source of truth
• Admin CSV override when the feed lags or breaks — the scoring engine replays from the upload
• A 50-pick memory cap in the live store keeps long drafts light on mobile

Built like a real product. The plumbing users never see is the plumbing that decides whether this thing survives.

• 347 Pest feature and unit tests, plus 24 Vitest tests on the JavaScript stores and offline storage
• PHPStan static analysis at a meaningful level, not level 0
• Sentry error tracking on both the Laravel and client sides
• Passkeys and TOTP for two-factor, plus Google and Apple OAuth
• Postmark with RFC 8058 one-click unsubscribe and plain-text fallbacks
• A service worker, IndexedDB storage, and PWA install flow

Paying for the quiet months. A draft game has a brutal traffic shape — dead most of the year, mobbed for a few days in April, June, July, and October. I encoded the calendar into the deployment itself.

• Offseason: zero warm Lambdas, scheduler every 15 minutes
• Pre-draft: two warm instances, scheduler every five minutes
• Draft week: 10 warm instances, scheduler every minute

It's the kind of cost optimization a small company would do for real money. For me, it's what keeps the project inside the budget of a personal side project.