Real builds for real operations.

Lock sells sound kits to producers, and Shopify had become the wrong container for that: clunky to work in, limited in what the store could look like, and taking platform fees on every sale. The membership program he wanted to launch was something Shopify couldn't do well at all. So the site was rebuilt from scratch — Next.js on Vercel, Supabase behind it, Stripe handling payments — for full creative control and lower fees per sale, on a foundation the membership could actually run on.

That membership machinery is already built. When a payment clears, a Discord bot delivers the member role automatically; a daily job re-checks every subscription and revokes access if one lapses. Honest status: built, not yet live — the Members Area shows "Coming Soon" on the live site as of this writing.

This was also a threshold build in its own right — the first production e-commerce site I've shipped handling real customer and payment data. That raised the stakes on the cutover: real customers, real cards, no room for a broken checkout. It went out clean.

See it live — inthelockboxx.com

This is a personal build, not a client engagement — and the bootcamp it serves isn't for sale here. I'm developing an AI-fluency curriculum on my own time, and the War Room exists to keep it sharp while I do: a self-directed project built to practice agentic-workflow skills — the same kind of work I do in enterprise IT — against a real, evolving artifact instead of a toy problem.

Every night, the self-hosted n8n scheduler kicks off a run. A Claude-powered agent, wrapped in a FastAPI service, researches the AI-education market, critiques the current curriculum against what it finds, and writes concrete change proposals. Each stage streams through Supabase Realtime into a Next.js dashboard laid out as three monitors — so there's never a moment where the pipeline is doing something I can't see.

The design rule that makes it trustworthy: the agent never touches the live curriculum. Proposals land in a review queue, where every diff is read and approved or rejected by hand, and every version is snapshotted before a change applies — so anything can be traced back or rolled back. The autonomy is in the research and the drafting; the authority stays human.

The Local Website Program is the service model this site sells — and it went through the same treatment a client project would. Before pitching a single business, I researched the local market: what businesses around here actually pay for a website, what they get for it, and where they get stuck. The program was built and priced against that research first — the pitch came after the homework, not before.

The shape it took: a fully managed website on a simple monthly plan, with most builds carrying no upfront design bill; more complex builds — e-commerce, multiple locations, custom integrations — get scoped individually. Process automation stays a separate, request-only service: if a specific task is eating hours every week, it gets scoped on its own, never bundled in by default.

It's also the most personal project on this page. The program is the start of a path toward working for myself full-time, which means the model has to hold up in the real world, not just on paper. This site is the program running on itself.

Pack runs a trading card storefront — a family business selling across TCGPlayer, Wix, and eBay at once. Three channels meant three sets of daily orders, every one of them copied into the bookkeeping sheet by hand, field by field, every single day. The first automation took that job over, built by remoting into the business's own machine and automating the work in place — same sheet, same computer, no new platform to learn, no migration to sit through.

The second workflow handles pricing. An Apify scraper pulls live TCGPlayer secondary-market Pokémon prices every morning and applies them across the matching product catalog. On the storefront, each market price shows with a strikethrough next to Pack's own below-market price — so the deal isn't a claim, it's visible against the live market, every day.

Together, the two workflows gave back 3–5 hours a day that used to go to manual entry — the kind of build that typically costs $2,000 or more to have custom-made. And it was done the way this whole practice works: the same people still run the store, they just got their mornings back.