About
Websites, paid media, SEO, video, machine learning, security. Each one arrived because a client was waiting. Each one changed how I do the others.
/002 The short version
A book site, a hat that made a cat look like a lion, an importing business, a digital agency called Mane and a media company called Hummingbird. Machine learning and security came out of the client work and stayed.
Some of it sold. All of it taught me something expensive.
These days I build companies and the systems that run them. The long version is below, and it's the part that explains the short one.
The long version
/003 The argument
Jack of all trades? No. Master of none is the part everyone forgets to argue with.
Here's what actually happens. You jump into the deep end of a trade because someone is waiting on you. You push at the boundaries. You get obsessed, and you stay long enough that it stops being a skill and becomes a craft. Then the next one arrives, and it's easier than the first, because you've already learned how to learn.
The value isn't the number of trades. It's that the thinking transfers. A manufacturer's instinct for tolerances changes how you scope a build. A filmmaker's instinct for pacing changes how a homepage reads. A security mindset changes every assumption you make about a system.
And because I've done the work myself, not just scoped it, I can usually tell you which of yours to fix, which to automate and which to leave alone.
/004 Where I go deep
Twenty years of deciding what not to build, mostly with my own money at risk.
Every seat: the code, the design, the brief and the invoice at the end of it.
Search and social, judged on what an enquiry costs. A click has never paid anyone's rent.
Fifteen years of watching people optimise for an algorithm that had already changed.
Custom models, forecasting and the agents that now run parts of a business, built for clients rather than demonstrated in a deck.
Scrapers, scanners and vulnerability detection. It started with one client's worst week.
/005 The long version
The first thing anyone paid me to make was a catalogue for a local business, printed in the back of a suburban newspaper. I was seventeen. It ran, the business sold from it, and I haven't really stopped since.
I said yes before I knew how. It wasn't the last time.
I studied Law and Commerce, majoring in marketing and management. Halfway through I built SloppySeconds, Books That Got Around, so students could sell their textbooks to each other instead of donating them to the void.
Terrible pun. First real website. The pun has aged better than the code.
That was the year the web stopped being a subject and became the thing I actually thought about.
Then I had a ridiculous idea and gave it the only name it could have had. Lion Cat Hat. I found a manufacturer overseas, built the store and started writing the ads. You've cat to be kitten me right meow. I ran Australia's Next Top Cat while the country was watching Australia's Next Top Model. It went viral more than once and ended up on television.
I wrote the ads, answered the emails and packed the orders, most evenings, in a house that had run out of rooms without a box in it.
The manufacturing relationships turned out to be worth more than the hats. I spun those into a second business, helping other companies design, manufacture and import their own products. A considerably less funny business than the one that paid for it, and a much better one.
Here's the part that mattered, and I didn't understand it for years. The factory that sewed my hats would have sewn anyone's. Same pattern, same freight forwarder, same price list. Not long after, there were copycats on eBay. The only thing I owned was the reason anyone wanted one.
That's the whole lesson, and it's the same one AI is teaching everybody right now. The capability was never the advantage.
I graduated and found that out quickly, so I didn't.
I went back to what I'd been doing on the side the whole time: building small ventures, and helping businesses work out what they were doing online.
Some of those clients are still my clients. The longest has been with me for seventeen years. It's the part of this I'm most proud of. Anyone can win a client. Keeping one that long means being right more often than you're wrong.
There was a tobacco importing business: shisha molasses, custom packaging and distribution, with three tiers of restricted licensing before a single carton could legally move. There were standalone apps. There were things I built for one client and never sold again.
Some ran for years. All of them ended. Most of them on purpose.
Test it, launch it, then scale it or kill it. Killing something that works a bit, but not enough, is harder than killing something broken, and it's the decision most businesses avoid for about two years longer than they should. Most of what I know about running a business I learned from the ones I killed.
Test it, launch it, then scale it or kill it.
Most of that work eventually lived under one name. Web, design, strategy and later AI: custom models, training, automation, neural networks doing forecasting for businesses who'd never had a forecast worth reading.
A digital agency, mine, for the better part of a decade. That's where the 200+ comes from: startups, SMBs and enterprise, from two-person businesses through to ANZ.
Some of them wanted a website. Most of them wanted someone to tell them the truth about the one they already had, and the second job is the one worth paying for.
A client came to me wanting a video. I didn't make videos.
They said they didn't trust anyone else to do it. So I said yes, and then went and learned how.
Not entirely from zero. I'd been making films for years: school productions, church events, anything that would let me hold the camera. But there's a long way between that and work a business can put its name to.
It became Hummingbird Productions. Five people, cinema cameras, gimbals, drones, VFX, animation, the whole nine yards. We made work for serious businesses and, at one point, a music video that went viral and ended up on television overseas.
The reason it worked wasn't the gear. Production companies were charging an arm and a leg, which priced out exactly the startups and smaller businesses who needed good content the most. We made the good version affordable. That was the entire business.
Then COVID arrived and took the media work overnight. Production was my cashflow. Production was now illegal.
So I treated it as time off and learned Python.
One app a day. This was before AI could write any of it for you. You read the documentation, you sat there and you worked it out. Full immersion is the fastest way I know to learn anything, and it's uncomfortable by design.
That turned into machine learning, then predictive analysis and data science. I built and sold custom trading algorithms. I built a computer vision model that tried to predict Myers-Briggs type from a photograph and got to 58% accuracy. Not good. Well past chance, though, and enough to prove the prototype did what I'd said it would. Some experiments are worth running for the answer, not the product.
All of it went back into the agency, which is why the AI work here was built under commercial pressure for people paying for it, rather than as a product looking for a market.
One of my clients was hacked. Their database was taken and the people who had it were threatening them.
Nobody would help. The professional incident-response firms were far outside what they could pay, and the clock was running.
So I cleared the calendar and did nothing else. Fourteen hours a day, most of them spent not understanding what I was reading. Today a single prompt will explain most of it, and how each piece connects to the next. Then it was manuals and forums, and slowly working out that I couldn't understand one part until I understood three others.
Two weeks later I had the basics of Kali Linux and one objective: get my client's data back.
I got it back, and handed everything I'd found to the authorities.
That's where the security work started. I still build scrapers, scanners and vulnerability detectors for clients, because the most useful thing you can do for a business is see it the way an attacker does. Nearly everyone I've done that for assumed they were too small to be worth attacking. None of them were targeted. They were scanned, which is not the same thing and is much cheaper for whoever's doing it.
/006 Now
I'm Digital Director at The Creative Works in Melbourne. That was a deliberate choice rather than a landing. After twenty years of running my own businesses, the interesting problem stopped being can I do this and became how much can this do, and for how many.
I still build my own. And I take on three engagements a quarter: advisory, builds and growth, for founders. More of the builds each quarter are AI agents, each running one part of a business around the clock, with a person signing off on anything that touches money or a client.
The jobs they do are ones I did myself first: the follow-ups, the inbox, the reporting, the 2am fix. It's the work that needed all twenty years before it.
/007 Why I write
I've made most of the expensive mistakes myself, in my own companies, with my own money. Which is why I'm quick to spot yours.
I write about what I worked out, so it costs you less than it cost me. That's the whole reason the notes exist.
Read the notes/008 Next steps
Advisory, Build or Growth, three engagements a quarter. I answer my own mail. If it's a problem I've had, I'll say so. If it isn't, I'll say that too.
Tell me what's in the way
