How we build digital products today
AI agents in development—tools in the team, not a show
For us, AI agents aren't a single writing tool. They run with the work from discovery through documentation: clear tasks, small steps, human review before moving on.
Iterative agent loops across the full development process
Agents don't replace a team. They extend it.
Good digital products don’t come from automation alone. Agents analyse, structure, and prepare work—with a clear goal, solid context, and people accountable for the outcome.
Our human-in-the-loop principle:
Working principle: Direction, Loop, Review
Direction
People set the goal, context, and priority. Problem, constraints, and a measurable outcome first.
Loop
Agents take on structured prep work in bounded iterations—analyse, design, implement, test, document.
Review
People review from a domain and technical angle. What holds goes into code, docs, and the next iteration.
Hypothesis first, then the spec — then the agent
AI agents can build software much faster. That also raises the risk of turning the wrong assumptions into working code more quickly.
That is why we combine hypothesis-driven product development with spec-driven development. The hypothesis describes what we want to achieve and learn. The spec turns that goal into a clear, verifiable brief for people and agents.
How are you using AI in development?
On the call: direction, iterative loops, and human accountability—not more code at any cost.
Clarity over buzzwords
Topics that either trip projects up early—or help them find their footing.
