
Accomplishments
1,700+
Issues & Pull Requests
Every significant change tied to a tracked item on the project board.
9,100+
Automated Tests
Passing across the app, its API and its domain packages (9,180 as of 18 November 2026).
24
Architecture Decisions
Each recorded as a formal decision record, with its context, the decision and its consequences.
In one of life's strange turns, I found myself running an 8-tap craft beer tasting room. Tracking the tap pipeline in the paper notebook I inherited was an exercise in futility and frustration. Nearly every taproom operator I spoke with did the same: a messy Excel spreadsheet or, more often, a paper notebook, even though most of them used tools for menus and social media. Taprooms have plenty of help with the front of house, but very few tools for the back-of-house work that matters just as much. HoppyTRM was born to save my sanity, and to help taproom operators spend less time in spreadsheets and more time pouring beer.
HoppyTRM is back-of-house software for taprooms, beer bars and tasting rooms that pour other people's beer. It follows every keg from order, to cooler, to tap, to the vendor picking up the empty, so you always know what's pouring, what's next on every line, and which deposits you're still owed. Today that means a Tap Board and Tap Pipeline with one-step Advance when a keg kicks; orders by supplier, tracked from pending to delivered, with order emails drafted from your own templates; empties tracked through vendor retrieval, with a pickup notice ready to send; reports on days to kick and deposits held, plus sales, gross profit and pour cost estimated from each keg's cost and pour prices until POS sales import arrives; and events, keg history, a beer catalog and supplier records. AI features such as Quick Add are in early access, and you review what they find before anything is saved.
Built by a taproom operator, HoppyTRM has run an 8-tap tasting room every day since September 2026. I'm building it with Claude Code as my development partner, working from the product specification, user stories and recorded architecture decisions kept with the project. HoppyTRM is in private alpha on desktop, tablet and phone. Request access at hoppytrm.com.
HoppyTRM
2026
|
Self







HoppyTRM Carousel
A look inside HoppyTRM, back-of-house software for taprooms, beer bars and tasting rooms that pour other people’s beer.
Swipe through to see:
🍺 The Tap Board and Tap Pipeline: what’s pouring and what’s next on every line
✨ AI Features: Quick Add, Enhance Records and Import Records, with your review before anything is saved
📋 Orders by supplier, from pending to delivered
🚚 Empties tracked until the vendor picks them up
📊 Reports on days to kick and deposits, plus estimated sales, gross profit and pour cost
📺 Digital Menus for the screens behind the bar (early access)
📚 Inventory, Events, Keg History, the Beer Catalog and more
Built by a taproom operator, HoppyTRM has run an 8-tap tasting room every day since September 2026. It’s in private alpha on desktop, tablet and phone.
Run a taproom? Request access at hoppytrm.com.
Screens show a sample taproom with example data.


A Different Approach to Taproom Management
I built HoppyTRM around the physical keg instead of the beer. The beer is what our business is about. It's what we pour and what our customers drink. But it's the keg that gets ordered, stored, tapped, emptied and picked up again. Each keg moves through nine stages on its way through the taproom, from Planning and Ordered to On Tap, Empty, Awaiting Retrieval and Retrieved.
Tracking the keg lets the app keep the beer separate from where each keg is and what state it's in, and that makes everyday work simpler and more flexible. On Deck, for example, is just the first keg in a tap's queue. It isn't a spot in the cooler or a separate menu. New kegs join the queue oldest first, a reorder made by staff is never undone, and putting the next keg on tap takes one step from the Tap Board. Taprooms exist to pour beer, but they run on kegs. Building around that means less time spent tracking kegs in notebooks and spreadsheets, and more time pouring the beer customers love.
Designed Around the Real World
A taproom is a hard place to use software: wet hands, a line at the bar and no time to dig through menus. So before any code gets written, I set the principles every piece of work has to follow. They build on standard UI/UX practice and adapt it to life behind the bar, and with AI doing much of the building, they're the first thing every session reads.
Fast, forgiving and easy. Someone with wet hands in the middle of a rush should be able to get things done quickly. When they make a mistake, they fix it in one step on the screen where they notice it, without a trip to another section.
Every screen has one job. A feature gets added only if it helps a taproom understand, manage or automate the flow of beer. Screens don't repeat each other, so each one earns its place.
There's always a way back. Anything that takes you to another page gives you a one-tap way back to where you came from.
Options without overwhelm. Every taproom is different. We prioritize different things and work under different constraints, so software that serves many taprooms has to be flexible and highly configurable. The challenge is offering all those options without burying people in them.
Let the tool do its job. When you split wood, you let the weight of the axe do the work. Forcing it only makes the swing less effective. Software is the same. A taproom owner shouldn't spend time on data entry or tracking keg deposits by hand. The software can do that, which leaves more time for finding and pouring great beer.
These principles aren't fully in place yet, and I've learned a lot along the way. I've never been a fan of "move fast and break things," especially in healthcare. But for HoppyTRM, I've focused on building the features a taproom needs to operate, and I've often accepted UI/UX I know isn't good enough in order to ship them. That's uncomfortable for me as a designer. Because HoppyTRM is still in private alpha, there's time to fix it, and I plan to bring the UI/UX up to my standards before it's released publicly.

Directing AI Like a Product Team
Principles only help if every session follows them, so I direct the AI agents the way I'd lead a product team. A shared instruction file gives every agent the same product goals, the domain rules it can't reopen without my decision, and the working agreements everyone follows. Four rules do most of the work.
Check the facts first. Agents read the live code and Git history before trusting any status document or handoff note, and they try a feature in the running app before describing it. Notes go out of date. The code doesn't.
Protect the data. A taproom's record of what it bought, poured and returned is never rewritten. Corrections add new records instead of deleting old ones, and bad data is refused with a way to fix it, never quietly repaired.
I make the product decisions, before any code. Navigation, scope, color, type and wording are settled with me first. Visual changes need mockups at phone, tablet and desktop widths, with a WCAG AA accessibility check, and I approve them before any code changes. The design system lives in one Claude artifact that every session works from. If a change grows past its estimate, the agent stops and checks with me.
Prove it works. Every fix starts with a test that fails because of the bug. Then the full build, typecheck and test suite have to pass, and the change gets tried in the running app at all three widths, in light and dark mode, and with the keyboard alone.
Most of these rules came from things that went wrong. A UI/UX audit in September 2026 traced many defects to one pattern built several different ways, and to fixes that repaired one screen while breaking its neighbors. Parallel AI sessions have filed duplicate issues, and I once stopped a merge because nobody had tried the change in the running app. Each time, the lesson went into the rules.

Key Accomplishments
1,700+
Issues & Pull Requests
Every significant change tied to a tracked item on the project board.
Every significant change tied to a tracked item on the project board.
9,100+
Automated Tests
Passing across the app, its API and its domain packages (9,180 as of 18 November 2026).
Passing across the app, its API and its domain packages (9,180 as of 18 November 2026).
24
Architecture Decisions
Each recorded as a formal decision record, with its context, the decision and its consequences.
Each recorded as a formal decision record, with its context, the decision and its consequences.
Tags
Branding, Creative Direction, Design, Product, Product Management, Software, UI/UX, Visual Identity, WCAG
Website
Company
Self
Role
Founder
Year(s)
2026
Location
Pasadena, CA, USA
Industry
N/A
Biz Model
Personal Projects

