AI Isn't a Crutch for Bad Developers ... It's the Unlock for Neurodivergent Ones

I was sitting at my desk at 10:47 PM on a Tuesday, Cursor open in front of me, when I finally understood what had been happening for months. My hands weren't moving. The cursor was blinking. And yet I was in flow ... the kind of deep focus I'd spent my entire career chasing and losing and chasing again. For the first time, my brain and the screen were moving at the same speed.

A major global study by the Tavistock Institute surveyed over 2,000 tech employees across companies like Colt, Nokia, Samsung, and Vodafone. Of the 2,176 respondents, 562 self-identified as neurodivergent ... roughly 1 in 4. The real finding was about the gap in disclosure: over half hadn't told their employers because they lacked a formal diagnosis or didn't see the value. Many developers haven't had language for what they've been experiencing. (Source)

The loudest voices in engineering keep framing AI as the great equalizer that will let mediocre developers ship junk code. They're missing what's actually happening.

AI isn't a crutch for bad developers.

It's the unlock for neurodivergent ones.

I've built a career on the assumption that the eight hours between "I know what to do" and "I've done it" was a tax I had to pay forever.

The Tax I Thought Was Permanent

I've been coding since I was thirteen. I have ADHD. I went undiagnosed until I was thirty-six.

Those twenty-three years were ... frustrating is too small a word. I never knew why getting pulled away from things made me so irritable. Why I couldn't get into the zone the same way other people seemed to. Why, once I finally found that zone, it felt like a physical injury when someone interrupted it.

I built a career on the assumption that the eight hours between "I know what to do" and "I've done it" was a tax I had to pay forever. That the friction was just ... me. That the forty-five minutes of warmup before I could write the line I'd been holding in my head all morning was a character flaw I needed to engineer around.

The hardest part of writing code with ADHD was never the thinking. It was the typing. The context switching. The rebuilding of state every time I came back to the terminal after lunch.

My brain moves like a conductor thinking about thirty things at once. The strings need to come in here, but also watch the percussion section, and is that soloist rushing, and what's happening with the dynamics in the back rows? When you're writing code ... one file, one function, one line at a time in a linear text editor ... that multithreaded mind becomes a liability. I would get stalled because I would just be continually thinking and rethinking and rethinking. Five ways to implement something before I even started typing.

That tax is gone.

When the Conductor Found an Orchestra

It was when I realized that the state of flow you can get with AI is almost a benefit of my focus, not a workaround for it. I can hyper-focus on certain things. The ability to get from zero to one and then continue to iterate allows your brain and the AI to move almost as fast as your brain moves for the first time in my entire life.

I've never thought AI was cheating. One of my top five CliftonStrengths is Competition. There's a saying in sports: if you aren't cheating, you aren't trying. The best developers find the cheat codes and the holes. That's what makes them great. They think through edge cases. They approach things differently than people who follow the rules.

AI doesn't replace my judgment. It doesn't replace my ability to see thirty things at once and hold the architecture in my head. It eats the overhead for breakfast. The transcription tax. The syntax lookup. The "how do I express this thought in code" friction that used to burn forty-five minutes every single session.

AI doesn't replace my judgment. It eats the overhead for breakfast.

The Team I Didn't Design For

I've watched the same thing happen to engineers on my team who learn differently, think differently, work differently.

We assume everyone thinks like us. Even when we intellectually know better ... our day-to-day frustrations betray us. We build processes for carbon copies. We put people in boxes, and if they don't fit, we tell them to conform to the box, not to expand the box.

The truth is that we all have different ways of learning. Some learn orally. Some learn better visually. Some learn better by sitting in a classroom talking about theory. Others learn better by getting into the mess and making mistakes.

AI doesn't fit into a perfect mold, and the unlocks are for everyone differently. How we approach AI is different, and that's actually the beauty of it.

One person on my team has a pilot's license. They think in checklists and systems. For them, plan mode is the unlock. Being able to spec every last thing out, have the AI ask clarifying questions, keep up with their brain wave ... that's the fit.

Another has a scholar background. Master's degree, deep researcher. They gravitate toward ask mode. Starting with questions, iterating on understanding before building. The AI keeps up with their need to explore.

A third just combos their way in and goes straight toward agent mode. No spec. Just velocity and iteration.

Iron against iron creates sparks. That's the type of sparks we need on our teams. But for years, only one of these thinkers got to operate at full speed. The task-list person got buried in planning paralysis. The scholar got stuck in research loops. The combo-coder got dinged for "moving too fast without thinking."

The neurotypical engineer who could just sit and grind for six hours straight ... that person set the pace. That person's workflow was "normal." Everyone else was a deviation that needed accommodation.

Now the ones who used to get edged out are shipping faster than that engineer ever did.

The ones who used to get edged out are shipping faster than that engineer ever did.

The Tax on Focus

A 2023 controlled experiment published in arXiv found developers using GitHub Copilot completed tasks 55.8% faster than those without AI assistance. The researchers didn't segment by neurotype, but the mechanism is clear. AI reduces activation cost ... the cognitive tax of starting. For people with ADHD, that tax is higher. (Source) The distractions of the modern-day world ... Slack, email, calendar notifications ... hit different when your executive function is already a limited resource.

The AI doesn't just make me faster. It makes the fragments of focus I still have more valuable. When I get twenty minutes between meetings, I don't spend nineteen of them rebuilding context. I spend one. Then I ship.

This is why the "AI is for bad developers" crowd misses what's actually happening.

They're so busy defending the keyboard that they're missing who just got let into the room.

They're not wrong that AI can paper over gaps. A junior engineer can ship senior-looking code without understanding it. A developer can generate features without ever sitting with the system long enough to develop judgment. Those are real risks.

But they're wrong about who the tool is unlocking. They're wrong about what "good developer" even means when the constraints change.

Different brains were always doing the work. They just had a worse tax rate.

The engineers who could sit and grind six hours straight weren't better. They were better suited to a toolset designed by people who could sit and grind six hours straight. The IDE, the terminal, the git workflow, the PR review process ... all of it optimized for a particular cognitive style. The rest of us were paying a productivity tax for the crime of thinking differently.

AI doesn't level the playing field by lowering the bar. It levels the playing field by removing a barrier that was never about skill in the first place.

What This Means for Leaders

I keep watching leaders who think their job is policing AI usage. Tracking credits, enforcing policies, making sure people aren't "cheating." They're missing the same thing I had to learn about open floor plans destroying deep work. The same thing I had to recognize about "culture fit" being code for "thinks like me."

We've gotten better over the years. More aware of different learning styles, sensory processing, the invisible tax of environments that don't fit. But we still default to designing for the median and accommodating the edges.

AI is flipping that. The edges are finding tools that fit their minds for the first time. The median is discovering that their six-hour grind sessions aren't actually the only way to ship quality code.

Your job as a leader isn't to police who's using AI and how. It's to notice who's suddenly shipping faster, who's suddenly contributing more, who's suddenly engaged in ways they weren't before ... and ask what changed.

Sometimes the answer is a new tool. Sometimes the answer is that the tool finally fits the person.

Different brains were always doing the work. They just had a worse tax rate.

The tax is gone.

One email a week from The Builder's Leader. The frameworks, the blind spots, and the conversations most leaders avoid. Subscribe for free.