The flow state AI stole

Jonathan Beard's two recent posts (one, two) really resonate with me. I use Claude daily for work and personal projects. The cognitive dissonance is deafening as I use AI to do things I've never been able to do before, all while the joy in solving hard problems disappears.

That last twenty percent is where the engineer used to live. It was never the fun part, but it was the formative part, because it was the part that forced contact with the substrate and built the judgment that contact produces.

Which reminds me of the conditions required to be in a state of flow, especially number 3. The struggle Beard describes is exactly the challenge/skill balance from condition 3. For me, the entire reason I even learned to code was to chase the flow state. It's when I've learned the most and felt the most joy when coding. I generally call this being nerd sniped w/r/t coding (for me the ultimate flow state is casting dry flies at rising trout!).

1. The activity must have clear goals and progress. This establishes structure and direction.
2. The task must provide clear and immediate feedback. This helps to negotiate any changing demands and allows adjusting performance to maintain the flow state.
3. Good balance is required between the perceived challenges of the task and one's perceived skills. Confidence in the ability to complete the task is required.

I think it's fair to state that well structured sessions with AI satisfy 1 and 2. 2 might almost be too immediate sometimes. The problem is 3: when 1 and 2 come at you this fast, the constant interruptions from AI crowd out the deep work required to even feel 3. I don't know about you, but I've never felt myself to be in a flow state when coding with AI.

The concerning part is all the conversations around "what about the junior engineers" and continued skill development for all engineers? Which is a valid and real question that I've yet to see a good answer to in this wave of AI adoption. Individually I think there are solutions. Don't offload all your thinking to AI. When you run into something you don't know, ask AI to teach it to you. Yet the conversations at large around adopting AI focus on 10x-ing productivity, and often leave out the time we'd otherwise spend learning something new.

That pressure cuts deeper than lost learning time. AI doesn't actually remove the work. The workload creeps into other parts of the day, shifting effort from producing toward reviewing and understanding ever more output. The productivity fatigue accumulates as we continue to adopt more AI tools.