Blog
We stopped selling work we were good at. Here's what it taught us.
In July we stopped selling several things we were genuinely good at. Hardware and firmware work, connected sensor projects, 3D scanning and printing, industrial design. We had built real capability in all of it and we had clients who valued it. Deciding to stop was harder than any technical problem we have solved, and I think about it every time we ask a client to retire something. Here is what it
CodeFish is now listed in South Australia's AI Capability Directory
CodeFish Studio is now listed in the AI Capability Directory SA , run by the South Australian Department of State Development. The directory exists to answer a simple question: if you are a South Australian business that wants to use AI, who here can actually do the work? It lists local organisations by what they deliver, which technologies they work with and which parts of a business they help. F
Six systems, one business, and no single version of the truth
Ask two people in the same business how many jobs went out last month and you will often get two different numbers. Both of them are right, which is the interesting part. One is counting what got dispatched. The other is counting what got invoiced. The dispatch number lives in a scheduling tool, the invoice number lives in the accounting system, and nothing sits between them to say which question
You and your competitor are paying the same vendor to run the same business
There is a line item in your accounts that goes out every month without anyone thinking about it. Somewhere across town, a business you lose tenders to has the same line item, for the same platform, at roughly the same price. Between you, you are funding the same roadmap. Whatever that vendor builds next, you both get. Whatever they decide not to build, neither of you gets. You are paying, in para
Design discipline costs less, and presents better
Every piece you bought was nice. The chair, the rug, the lamp. Then it all landed in the same room and nothing went together, because nobody decided what the room was before they started buying. Software fills up the same way. And the fix most teams reach for, adding more, is the most expensive one on the table. It rarely shows up as a line item. It shows up as the same argument every second sprin
Your last three software projects didn't stick. The model is why.
Most businesses we meet have already bought software. Often three or four times. There is a portal somebody built in 2019 that two people use, a workflow tool that got abandoned during the rollout, and a dashboard nobody opens because the numbers in it stopped being true. The instinct is to blame the build. Wrong developer, wrong platform, wrong spec. Sometimes that is fair. Far more often the sof
Your roadmap is losing to the day job
We get shown a lot of plans. Most of them are sound. Someone has thought hard about what the business needs to change, put it in a deck, and got agreement in the room. Twelve to eighteen months later, when I ask how it went, the answer has the same shape almost every time: we got partway, then things got busy. Nobody made a decision to stop. That is what makes this hard to see coming. Think about
What South Australia's Royal Commission into AI means for you
On 10/08/2026, Premier Peter Malinauskas announced a Royal Commission into Artificial Intelligence — the first anywhere in Australia. It sits on a $3 million budget, will be run by three commissioners chosen through an international search, starts on 01/10/2026 and reports by 01/07/2027. That is nine months of hearings into how AI is used in this state, and most of the businesses that will be affe