Future
Where we're heading, and what we're building next.
AI, Vibe coding, coding agents
We fully support the use of responsible AI in the development lifecycle. There are some things it does very well, and does quickly. It is a tool. It is not a solution (at least not today). Any tool requires a responsible operator, and we have learned what it takes to be a responsible operator. It was a long, hard road, with it’s fair share of dissapointment and failures along the way. But today, we can utilize the strengths of AI and at the same time ensure that quality, security, focus and requirements are always monitored and maintained. We do not have AI agents committing, pushing or deploying code that has not been properly reviewed, understood and validated with testing, both automated and manual.
We see a very narrow window of responsible AI usage. To the “left” of the window is the universe that believes anyone can write a prompt and build a complex, reliable, cost-effective, secure, maintainable and enhanceable system, and to the “right” is the universe that believes AI is useless and untrustworthy and has no place in software engineering.
To the “left”, we wish them luck. We have experience with taking over projects that were Vibe-coded and the results are almost universally identical – pretty much everything is thrown out and we start over. The end result is usually a longer and more painful process than just doing it right from the start. The best recent analogy is using something like WebMD or AI to diagnose an issue and then going to the doctor and trying to convince them that you know what’s wrong. No, you probably don’t and should never assume web content or AI generated output is superior to true knowledge, experience and skill.
To the “right”, we ask, do you use an IDE to develop code, or a compiler, or any other of the modern tools or technnologies that help you do a better job, or do it faster and cheaper? Do you use frameworks like React or Vue, or development stacks like .Net? We are sure you do. AI is no different. Learn to use it responsibly and in a controlled and curated process.
Believing a cadre of coding agents working asynchronously for a day is going to produce a quality project is essentially believing that nine women can have a baby in one month instead of one woman requiring nine months.
The real work must be done, and while we can maximize the use of tools to improve our work, the responsibility ultimately lies in someone’s hands, and we are eager and prepared to be those hands.
Our own products
Alongside client work, we often build products of our own. Check out Gym Journal if you like.
Research
We keep exploring: new platforms, new kinds of interfaces, and problems that haven’t been solved well yet. Some of that work becomes products, some becomes expertise we bring to client projects.