Agentic Programming and the Neurodivergent Brain
I've been reflecting on why an agentic pair programmer works so well for my neurodivergent brain — and for me it has to do with speed, but most of all with cognitive fit.
Most of my work lives at the level of systems, flows, invariants, and failure modes. The hardest part is not writing code; it's holding intent, structure, and consequences in my head at the same time.
An agentic pair programmer quietly removes friction in exactly those places.
Where the Friction Goes
It externalizes executive function
State, sequence, and "what done looks like" are held outside my head — without judgment or social overhead. That reduces working-memory pressure immediately.
It turns intent into concrete artifacts
Architectural ideas become code, structure, and glue without losing momentum in translation. The mechanics stop competing with the thinking.
It provides continuous feedback without breaking flow
No batching, no meetings, no context resets. Just tight loops: try → inspect → adjust.
It's socially flat
No hierarchy, no ego, no performative alignment. I can think out loud, change direction mid-stream, or dive deep into edge cases without friction.
It adapts to nonlinear thinking
Deep dives, abstraction jumps, obsessive detail, system-level synthesis — the agent follows the mode switch instead of resisting it.
What This Actually Means
Most importantly, it lets me stay focused on intent, correctness, and system behavior, while the mechanical translation happens alongside me.
This doesn't feel like "AI replacing programming."
It feels like cognitive infrastructure — an extension of the same impulse that leads me to build generators, workflows, explicit models, and observable systems in the first place.
Not a new way of working.
A missing piece that finally fits.
Originally posted on LinkedIn.