A company that runs itself

The founding spec for automating Lumi Fave, and why trading cards are the right substrate

This year I invested in Lumi Fave, my brother's company. On my timeline that's one line. This is what's behind it.

The plan is not to run a shop. The plan is to build a company that runs itself, and to use a real, physical, revenue-generating business as the testbed.

The division of labour is clean. My brother runs procurement, logistics and customer support. I build the machine around it: operations, accounting, marketplaces, IT, content. There is no going concern to retrofit: when I signed, the shop was still unpublished, so the machine gets built in parallel with the catalogue instead of bolted onto it. Every process we touch gets modelled as a service; human judgment stays only where it is actually worth something: deciding what to buy and whether to ship. Everything around those decisions is plumbing, and plumbing can run unattended.

Why trading cards

We picked the niche half by spreadsheet, half by biography: two brothers who grew up with Magic and Pokémon. But a business you want to automate needs a product that cooperates:

TCG accessories hit every criterion at once. Cards, graded slabs and sealed boxes all have dimensions that have been standardised for decades, so an acrylic display case designed today sells unchanged in ten years, and the catalogue compounds instead of churning. The products are compact and durable, which makes logistics a solved problem. And the demand lives entirely on digital marketplaces with real APIs.

Two brands, one machine

Cardmuseum is the demand side: the shop where collectors buy sealed product and accessories (Shield N Show next to GradedGuard and VaultX) and where every marketplace channel converges. Shield N Show is the margin side: our own accessory line, led by acrylic display cases for booster boxes and graded slabs, plus holders, sleeves and binders, sold in the shop, direct, and wholesale to retail partners. That’s the old shovels-in-a-gold-rush play, with one addition: we also own the shop that sells them.

What counts as proof

Anyone can draw a convincing architecture diagram. So the proof will be the P&L, and the hours. If the thesis holds, revenue grows while the human hours per week shrink, until running the company is a four-hour workweek of pressing approval buttons.

The stack starts small (Shopify, n8n, Google Sheets, a handful of Cloudflare Workers), and every workflow we add has to pay for itself by deleting a recurring task. I’ll publish the methodology and the numbers as we go.

If it works, you’ll read it here. If it doesn’t, you’ll read that too.