AI has shortened the distance between my thoughts and working software. A recent interview reminded me that explaining those thoughts still requires a deliberate shift.
One reason I have enjoyed coding with AI so much over the past year is that I can move directly from an idea to something I can build and test.
I mostly work with my AI agents through voice. I talk through a problem, explore possibilities, change direction—sometimes simply give them a brain dump. They help turn that into something structured and actionable.
I don’t have to organise every thought before getting started. I can stay focused on the problem.
The same applies when I work with other engineers. I use AI to help turn my working thoughts into documentation, explanations and handovers. I ask the agents to consider the reader’s role, what they need from me and which details matter. I still review the result, but I no longer have to do all that translation from scratch.
The communication work hasn’t disappeared. AI has reduced how much conscious effort it takes.
A recent interview reminded me of the difference.
I was speaking with fellow engineers who needed to understand how I approached problems—not just whether I could deliver a solution. They needed me to explain assumptions, discuss trade-offs and make my judgement visible.
Instead, I sometimes stayed in builder mode: thinking ahead to implementation while leaving too much of the reasoning unspoken.
My confidence in working through architecture options with AI also made me too quick to treat some questions as straightforward. I wasn’t intending to dismiss them, but I overlooked their purpose: to understand how I evaluate options and take responsibility for a design.
That was my responsibility. The interviewers could only assess what I communicated, not what I believed I could work out once I started building.
I hadn’t interviewed for a long time. I had also become accustomed to having AI help bridge the gap between my thoughts and someone else’s expectations. In the interview, I needed to make that shift myself.
This hasn’t made me any less enthusiastic about AI-assisted development. Reducing communication overhead is part of what makes it so useful: I can spend more time exploring ideas, testing assumptions and delivering software.
But it has given me a useful reminder.
When I’m building, the work can reveal my thinking. When I’m meeting people for the first time, I need to bring that thinking into the conversation.
Next time, I’ll remember to switch modes.
Leave a Comment