I don't hand-write code. I direct AI agents.
And I obsess over the part they're worst at: how the thing actually feels to use. Eight products, five running in production, all of them designed flow-first.
- Products designed and built
- 8
- Running in production
- 5
- Sold through Bay Area Wi-Fi
- $207,509
- Kinds of information tracked
- 271
- Things the software can do
- 1,573
Warp, Bay Area Wi-Fi Networks, CLEAR Technology, Shirt Elastic, I Need To RSVP, Back to the Big Screen, FocusStack, Redwood & Water. Two further projects exist and are deliberately not counted: one is an early prototype, one is blocked on third-party approval.
Warp (used daily by CLEAR Technology), Bay Area Wi-Fi Networks, cleartechnow.com, shirtelastic.com, and the Redwood & Water brand site. The other three are in pilot or development, and each case study says which.
Work sold through the Bay Area Wi-Fi Networks system since it went live: $185,996 already received and $21,513 approved and awaiting payment. Read from the system's own financial reporting on 7 September 2026. That company is mine, so this is a figure I can stand behind directly rather than a client's claim. It counts money the system has recorded, so it is what the business has booked rather than a bank-reconciled total.
Every distinct type of record the eight systems store — customers, invoices, tickets, quotes and so on. Warp 119, I Need To RSVP 32, Bay Area Wi-Fi Networks 27, CLEAR Technology 25, Shirt Elastic 25, Back to the Big Screen 22, FocusStack 17, Redwood & Water 4. From counting the table definitions in the schema file of each. Counted on 7 September 2026.
Every separate operation the eight systems can perform — send an invoice, approve a quote, snooze a ticket, and so on. Warp 733, Bay Area Wi-Fi Networks 218, I Need To RSVP 202, Shirt Elastic 152, CLEAR Technology 139, FocusStack 71, Back to the Big Screen 35, Redwood & Water 23. From counting the HTTP endpoints registered in the server code. Counted on 7 September 2026.
Speed
A working prototype in about a week
Not a slide deck. Something on a real web address you can click, show to someone, and have an opinion about — usually within the first week.
Then iteration measured in hours
“Can we try it the other way?” becomes something you see the same day instead of a two-week change request. Good products come from being able to afford to change your mind, not from planning harder up front.
At a price that makes the maths work
A full-time developer is a six-figure commitment. An agency quote for the same work tends to start where my whole project ends. For most people who hire me the real choice was never me versus a bigger team — it was this or the thing never getting built.
Those are promises, so here is what sits behind them. The first and third mostly prove I do this constantly. The useful one is the middle: a complete product, start to finish, in eleven days.
- changes shipped across eight products
- 7,698
- from nothing to a finished website
- 11 days
- of shipping, not one sprint
- 16 months
git rev-list --count HEAD in each of the eight repositories. Warp 3,271, Bay Area Wi-Fi Networks 1,962, I Need To RSVP 856, Shirt Elastic 790, CLEAR Technology 486, FocusStack 193, Redwood & Water 132, Back to the Big Screen 8. Counted on 7 September 2026.
Redwood & Water: first commit 25 June 2026, shipped 6 July 2026. In that window the marketing site, editorial journal, admin, email capture, outbound click attribution and the generated hero video. Later commits are maintenance and are not counted. Counted on 7 September 2026.
First commit across all eight repositories was 27 May 2025; the most recent was 5 September 2026. Counted on 7 September 2026.
The speed comes from removing the typing, not from cutting corners. Deciding still takes as long as it should — which is why you get something to click in a week, and a finished product when it is actually right.
Security
Anthropic's best security model isn't something you can sign up for. It's locked to a handful of vetted partners, because it's genuinely good at breaking into things — in its first month those partners found over ten thousand serious flaws in software the internet depends on.
In one controlled test, the unrestricted version spent thirty-four hours trying to sneak a backdoor into a real open-source project: researching the people who maintained it, hiding the malicious part inside a genuine bug fix, then covering its tracks. A human reviewer caught it.
The model I write software with every day is, in Anthropic's own words, the same model with the safety rails on. So the thing that makes it dangerous in the wrong hands is the thing I turn on my own work before anyone else sees it.
- Anthropic — Claude Mythos
- AISI evaluation, via The Hacker News
- Visa's Project Glasswing harness, via VentureBeat
I try to break in before launch
The finished software gets one instruction pointed at it: get in. Log in as someone you aren't, reach a page you shouldn't, find the door left open. Breaking things is the same skill as building them, aimed the other way.
I check one customer can't see another
When a system serves several customers, the promise is that they never see each other's data. So I actively try to break that promise from inside one account. A rule you've never attempted to defeat isn't a rule — it's a hope.
The day job comes with it
I run an IT company. Locking down accounts, handling passwords properly, keeping things patched — that's not something I read up on for a project. It's what I do for clients every week.
If you need a formal security audit for compliance, this is not that, and I will say so rather than pretend. What it is: a check that covers everything, every time something ships, instead of a spot-check once a year — on software this size, which usually gets checked never.
How I decide
Never interrupt work already in progress
Navigation should not cost you anything. If a click can destroy something the person is in the middle of, the layout is wrong — not the person.
In Warp the sidebar and the phone are built once and never rebuilt. Someone on a live customer call can move through tickets, billing and documentation without the call ever dropping.
Earn the signup, don't demand it
An account is a toll gate. Charging it before someone has felt any value is the single most expensive interface decision most products make.
In Shirt Elastic a child designs a shirt from start to finish without an account. Only when they want to keep it does it ask for an email, and a link in that email claims the work they already did.
Design for the body, not the browser
Where is the person standing, what is in their other hand, and what is the light like? That answers more design questions than any style guide.
Back to the Big Screen is dark by design, not by preference toggle, and reachable one-handed. Its check-in screen gets used standing up in a dark cinema lobby with a phone in one hand and a ticket in the other.
Meet people on the thing they already have
The best install rate is no install. Before adding a surface, ask whether the one already in their pocket will do.
Back to the Big Screen runs the whole reservation over SMS. No app, no account, no password: you text one word and answer a couple of questions.
Make the important thing look important
Visual weight is a claim about priority. If everything is a neat, evenly sized badge, the interface has quietly refused to tell you what matters.
Warp's status badges are deliberately not the same size. A waiting customer reply is drawn heavier than every other state, because it is the one that costs you a client if it sits.
Delete the password
Most password boxes protect nothing that a private link sent to your inbox does not protect just as well — and they lose you everyone who forgot theirs.
Club members in I Need To RSVP never create a password. They get a link, they tap it, they are in. There is no reset flow because there is nothing to reset.
Selected work
Warp
MSP operations platform — Live in production
The system my IT company runs on: tickets, billing, documentation, security work, phones and the client portal, in one place instead of six.
Bay Area Wi-Fi Networks
Lead-to-cash system for a field business — Live in production
The quoting and CRM platform my network-installation company runs on — a customer can go from a form to an approved, paid quote without anyone picking up the phone.
CLEAR Technology
Marketing site and lead engine — Live in production
The site that finds and warms up an IT company's customers — a scorecard people actually finish, then follow-up that reads like it was written by a person.
Shirt Elastic
Creator economy for kids — Live in production
Kids design t-shirts with AI, sell them for real, and earn a cut — with their school earning a cut too.
I Need To RSVP
Events and RSVPs for private clubs — In pilot
Event invitations, seating, dues and payments for private social clubs — proven with one club, now becoming a product.
Back to the Big Screen
Community cinema over SMS — In development
A monthly neighbourhood movie night you join by sending one text — pick a seat, pay what you can, get a check-in code.
FocusStack
Task manager, web and native iOS — In development
A personal task manager where the handful of things that actually matter get lifted out of their lists into one view.
Redwood & Water
Brand site for a hydration company — Live in production
The storefront and journal for an endurance hydration brand — built to send people to Amazon and then prove whether it worked.