Real systems, real numbers.

A closer look at what we have actually built and run, not mockups or pitches. Where we can share a real screenshot we have, and where a build is internal to a client's business we describe it plainly instead.

Personal builds

Our own products, live and in daily use.

Personal product · Live

Archivist: a collection that keeps its own records

A card collection tracker and deck builder for a physical trading card collection. Every card is logged with a live market value, so the collection’s total worth updates as prices move. A built-in deck builder pulls only cards actually owned and clearly marks anything still needed, so buying decisions are based on the real collection instead of memory.

Impact: Turns "what do I actually own" from a search through boxes into an instant lookup, and stops decks from being planned around cards that were sold, traded, or never owned in the first place.

Visit archivist.cards

Public tool · Live

Sheikah QR: a code generator with nothing to sign up for

A free web tool that generates QR codes and barcodes entirely in the visitor’s browser: links, WiFi credentials, contact cards, and more, plus common retail barcode formats. Nothing is sent to a server, no account is required, and the codes it produces never expire.

Impact: Removes the usual friction around QR tools, an account, a paywall, or a shortened link that can quietly die later, and returns a code that keeps working for good the moment it is generated.

Visit s-qr.tech

Personal system · Live

A personal assistant system built from the ground up

A private assistant system handling daily journaling with real memory, task and research tracking, calendar syncing, a home inventory, an inbox triage assistant, and a bookkeeping pipeline that turns emailed receipts into a categorized ledger. Built solo and run continuously since early 2026.

Impact: Over 11,000 automated runs with a 99.9 percent success rate, handling daily life and bookkeeping with no manual bookkeeping and no manual busywork behind it.

Retail operations for a two-location cannabis retailer

Four connected systems built for the same real client, running two storefronts. Client details stay private; what follows is what each system actually does.

Retail ops · Credits & reorder

Vendor credit reconciliation and reporting

A credit engine that reconciles vendor credits against purchases across both storefronts and sends a weekly report broken out by location, feeding a reorder process that used to run entirely from memory and spreadsheets. Reports land where the team already works, no new software to check.

Impact: Replaced a manual weekly reconciliation with a report the team can trust, delivered automatically instead of chased down.

Retail ops · Inventory

A morning alert that flags what belongs on the shelf

A daily report comparing what is on the sales floor against what is sitting in back-of-house storage at both locations. Anything with nothing on the shelf but stock in the back is flagged first, aged backstock is called out on its own, and promotional or non-sellable items are filtered out so the list stays useful.

Impact: Delivered to the floor team before the store opens every morning, so restocking is decided from a clear list instead of a walkthrough.

Retail ops · Compliance

Closing out empty inventory records automatically

State compliance tracking requires every inventory record to be formally closed out once it reaches zero, a step that is easy to let pile up. This system checks two independent systems, the compliance tracker and the point of sale, and only closes a record once both agree it is genuinely empty, then sends a same-day summary of what was closed.

Impact: Cleared a 23-record backlog the first time it ran, and has kept that backlog at zero automatically every morning since, with no staff time spent on it.

Retail ops · Menus

A branded product menu from a single chat command

Staff previously built an up-to-date, branded product menu by hand whenever one was needed. Now a single command in the team’s existing chat tool asks for a location, a category, and a brand, and returns a finished, multi-page PDF pulled from what is actually in stock right now, formatted the way the business already presents its menus.

Impact: Turns a menu request into a finished PDF in under a minute, built from real-time stock instead of a stale count.