Border plasma, and a busy month on Petal Components
Since the playground shipped we have pushed nine releases in about a month. A glowing new border effect, a searchable select that fought back, and a lesson about z-index I should have known.
Writing on building web products and small internet businesses.
Longer-form pieces on what I’m building and learning: Elixir and Phoenix, web development, and the realities of running products on the internet. Newest first.
Building a live playground for Petal Components
We built a place where you can see every Petal component, dial in your own colours and radius, and watch the whole library restyle live. All in LiveView.
Rebuilding Petal Pro for an AI-first world
Petal Pro started as a Phoenix boilerplate with auth and billing baked in. v4 rethinks it around how people actually build now, with an agent in the loop from the first commit.
How I actually build now
I don't think of myself as a full-time developer. I think of myself as a technologist who ships. Here's what AI changed about the day-to-day, and what it didn't.
Teaching AI assistants to write Phoenix UI
If an agent is going to write your HEEx, it should reach for real components instead of inventing Tailwind soup. That is why we built an MCP server for Petal Components.
Shadcn for Phoenix, and what 1.4 million downloads taught us
Petal Components is the open source UI library we wish we had had. The idea behind it, and what passing 1.4 million downloads showed us about what Phoenix developers actually want.
Designing multi-tenancy in Petal Pro
Most real SaaS apps are multi-tenant sooner than they expect. Here is how Petal Pro models organisations, memberships, and per-org billing so you do not have to invent it under deadline.
Killing our Alpine.js dependency
Petal Components used to lean on Alpine for interactivity. Moving that to Phoenix’s own LiveView.JS made the library lighter and more at home in a Phoenix app.
Moving Petal Components to Tailwind v4
The upgrade to Tailwind v4 touched every component we ship. What changed, and why the pc- prefix matters more than it looks.
What building Kindly during the pandemic taught me
We built a kindness marketplace in the middle of lockdown. It reached more than 20,000 volunteers and taught me more about product than any calm year could have.
From law to code
People assume I have always built software. I have not. The path from a Melbourne law office to shipping products people use every day was anything but straight.
Leading a small dev team without getting in the way
The hardest part of bringing on other developers wasn't the technical side. It was learning to hand something over and actually let go of it.
Building HealthVoice, an AI receptionist that knows when to hand over
A clinic's phone never stops, and the front desk can't be in two places at once. HealthVoice answers the calls. The hard part wasn't the AI, it was deciding when it should get out of the way.
My first app, built end to end
The Yoto Developer Challenge got me to ship a complete app on my own, back-end and all. I didn't win. I walked away with something I wanted far more.
Designing Doseloop dark-mode-first
Most health apps feel like a spreadsheet or a hospital. I wanted Doseloop to feel calm. Here is what designing it dark-mode-first taught me about front-end craft.
Building Talentbookr for a friend
A friend had a sharp idea for booking live entertainment and needed someone who could take it from a sketch to a real product. That someone was me.
What running a bootstrapped studio actually teaches you
Wicked Sites has never taken a dollar of outside money. Building it taught me every part of a business, usually because there was no one else to do it.
The most valuable hour of any project
The hour you spend scoping a project properly saves you weeks later. Here is how I run it, and why clients rarely see it as the most important thing we do together.
Turning our fridge chart into a Yoto app
A hand-drawn sticker chart for my son is becoming Chores + Rewards, my entry for the Yoto developer challenge. Here is what I'm setting out to build, and the two bets I'm making.
The last ten percent
The difference between an interface that works and one that feels finished is a hundred small decisions no one puts on a feature list. That last ten percent is most of the craft.
Hello, world
Why I built this site: a public record of what I make, what I ship, and what I learn along the way.