AI, Cognitive Load & TBI:

The Unlock

Some context first, because the narrative around AI and brain injury usually gets this wrong. My degree is mechanical engineering, and that's how I started — designing space systems, rockets, electro-mechanical hardware. Stuff you can hold.

About a decade ago I shifted into systems engineering. That move took the direct design work off my plate and replaced it with integration — how all the pieces fuse together, what happens at the interfaces, why a thing that works on its own breaks when it joins the rest of the system. Less hands-on building, more zoom-out thinking.

Six years ago I moved again, into software as a Product Owner and Program Manager. Another step away from designing the thing myself. I could spec a system, define what it should do, write the requirements — but I couldn't code one. Not really. I'd chip away at scripts that took me ten times longer than they should have, and lean on engineers for anything past that.

None of that gap had to do with brain injuries. It was the natural cost of three career chapters that pulled me further from hands-on building each time.

Then I got Claude Code. Suddenly the gap closed. Whole tool suites, compliance systems, dashboards, automation pipelines — stuff that would have taken a team of devs weeks, I'd prototype in a day. For the first time since I left hardware, I could actually build the thing instead of just describing it.

That's the unlock. Not a redemption arc. A systems thinker who finally got to ship software. The brain injury part comes next — not because AI fixed something concussions broke, but because AI's pace ran straight into a ceiling concussions left behind.

Then came the bill.

The Cost

The fog crept in. Slowly at first, then all at once. I'd have three or four AI sessions going, hopping between projects, working at a pace I'd never worked at before. And then I'd hit a wall.

I'd lose track of everything. Couldn't tell you what I was building in one window because my head was still in the last one. I'd have to stop, close it all, and gather myself just to make sense of what I was doing. I started calling it slopper syndrome — the fog, the sloppy thinking, the empty tank, that feeling your brain ran out of whatever it runs on.

It's not burnout. Burnout takes weeks. This hits in hours. Sometimes less. One session too many and the whole day's done.

What the Research Says About TBI and Cognitive Load

The science here is pretty clean. After a brain injury, your brain has fewer thinking resources to work with^[3]. Not worse strategy. Not laziness. Fewer resources. Picture a computer with less RAM — the operating system still runs, but it can't keep as many programs open before things lag.

One brain-imaging study found that TBI patients did just as well as healthy people when the task was easy^[2]. But when the task got harder — more to track, more to juggle — their brains couldn't dial up activity the same way. Working memory hit a ceiling the healthy group never touched. The researchers said the problems were "confined to the high cognitive load trials"^[2].

That's the whole game. Easy load: fine. Hard load: crash. Lower ceiling, steeper drop.

The rest of the research fills in the picture:

• Slower processing. One of the most common things you lose after a TBI is raw mental speed — researchers call it "highly vulnerable to brain damage"^[3]. The slower your brain processes, the more tired it feels^[4].
• Switching tasks costs more. One concussion study found switch costs were "significantly greater in the participants with concussion than in controls", and the gap stuck around for at least a month^[5]. The reason: damaged wiring between brain regions^[6].
• Mental fatigue shows up out of nowhere. Post-TBI fatigue "appears suddenly without warning during mental activity, sometimes after seemingly trivial activities"^[7]. You don't always see it coming.
• About 65% of moderate-to-severe TBI patients report long-term thinking problems, and up to 15% of mild TBI patients still have issues a year out^[3].

An injured brain works like a broken filter. A healthy brain takes in noise and quietly decides what matters and what to ignore. An injured brain can't filter the same way. More gets through. The system overloads faster.

AI Brain Fry: It's Not Just Me

In March 2026, Boston Consulting Group ran a study on 1,488 US workers and coined the term "AI brain fry" — the mental wipeout that hits when you're using or overseeing AI tools past what your brain can handle^[9][8].

Here's what they found^[9]:

• Productivity rises from 1 to 2 AI tools at once
• Productivity peaks at 3 tools
• Productivity drops past 4
• Workers doing heavier AI oversight reported +14% mental effort, +12% fatigue, +19% information overload, and +33% decision fatigue
• Software developers were one of the hardest-hit groups — 18% reported AI brain fry

And that's people with healthy brains. Their ceiling was 3. The cost was measurable.

With TBI — less processing power, harder task switching, a leaky filter — the ceiling has to be lower. My experience says it is. Mine is 2. One step below the healthy peak. Exactly where you'd expect it if you put the two pieces of research next to each other.

It's Not the AI. It's the Pace.

Let me be clear about one thing. AI isn't the problem. Claude Code is one of the most useful tools I've ever touched. It bridged the gap between knowing what to build and actually being able to build it. I'm not against AI. I'm just trying to be honest about what it does to your workload — especially if your workload runs through a brain that already has limits.

Here's what happens. AI builds stuff faster. When every task takes less time, you start more tasks. You stop waiting. You stop resting. You jump to the next idea because you can. The natural pauses that used to live between tasks — thinking time, transition time, just breathing — disappear.

Research on AI-assisted coding shows the same pattern. GitHub Copilot studies found that AI cuts the load for repetitive work but creates "cognitive debt" — every suggestion is a decision (accept, reject, change) and those decisions cost something even when they feel free^[11]. The net load depends on the task, not the tool.

On top of that, every time you switch between projects, your brain pays a tax. Researchers studying focus have shown it can take more than 20 minutes to fully get back into a task after an interruption. With TBI — damaged wiring between brain regions — that recovery is even longer^[6]. And AI work bumps up how often you switch.

Stack it all up: less time per task + more tasks + faster switching + fewer thinking resources = overload. The AI didn't cause the damage. The pace it unlocked just exposed the ceiling.

The Natural Experiment: Home vs. Work

I've accidentally run my own little experiment. Here's the setup:

The token limit at home is an outside-in brake on my brain. It does exactly what TBI rehab guides recommend: structured pacing with hard caps on mental work^[13]. I didn't design it that way. The pricing model just happens to give my brain the limit it needs.

At work, the access is unlimited. No external brake. Nobody tells you to stop. And if you're someone who spent years working around a coding gap and finally has a tool that closes it, the pull to keep going is strong. It's not laziness. It's the opposite. You push because you finally can. And pushing past the ceiling is where the damage happens.

The 2-Terminal Rule

Two windows is where I find the line between getting work done and keeping my brain intact. At two, I stay productive and I can keep track of what's happening in both. The wait time while Claude is thinking turns into a forced micro-break — a recovery gap I didn't plan but my brain needs.

At three, I can't finish things in any reasonable amount of time without losing brainpower. Context bleeds between windows. I forget what I was doing and why. My judgment drops. The fog rolls in.

Same principle as the Neuro-RPE scale I use for training. In the gym, I track muscular and neurological RPE and back off before symptoms force me to^[13]. At the keyboard, I cap the load before the crash instead of recovering from one. Autoregulation isn't just for the barbell.

The Gap Nobody Has Filled

Here's something that caught me off guard when I started digging: nobody is studying knowledge workers with TBI who use AI tools.

The research lives in two separate boxes. One box: decades of neuroscience on cognitive load, task-switching, and working memory after TBI^[2][3][5]. The other box: brand-new (March 2026) research on AI tools and workplace cognitive load^[9]. Nobody has connected the two.

No one has asked what happens when you give a brain with fewer thinking resources a tool that kills natural pacing, ramps up switching, and unlocks a workload that would tax a healthy brain. My experience sits right at that intersection. So does yours, if you read this and recognize yourself.

This matters because AI tools are about to be everywhere. If you have a brain injury and you work in tech, product, engineering, or any kind of knowledge work, you're going to be using AI whether you want to or not^[14]. Understanding how it hits your specific limits isn't optional. It's the job.

What I'd Recommend

Based on the research and my own field data, here's what I'd tell anyone working with TBI and AI tools:

• Find your number. Respect it. For healthy workers, the BCG study found a ceiling of 3 AI tools^[9]. With TBI, expect yours to be lower. Mine is 2. Yours might be 1, might be 2. Test it honestly. The ceiling isn't a failure. It's a fact about your brain.
• Build in forced breaks. AI kills the natural pauses between tasks. Put them back on purpose. A 10-to-15-minute break between things, even when you feel fine. The research shows TBI fatigue hits without warning^[7]. Don't wait for the crash to tell you it's time.
• Batch similar work. Every switch costs a TBI brain more than a healthy one^[5]. Instead of bouncing between three different projects, do one at a time. Let AI's speed help you go deep, not wide.
• Watch for decision fatigue. BCG found a 33% jump in decision fatigue from AI oversight^[9]. Every suggestion you accept, reject, or change is spending something. Notice the cost even when it feels free.
• Cap your deep-work hours. Set a hard limit on heavy AI work per day. Not screen time in general — focused, high-demand AI work specifically^[12]. Your ceiling will be lower than your coworkers'. That's fine. Quality work inside the window beats wrecked work outside it.
• Use outside limits whenever you can. Token budgets, timers, scheduled breaks — anything that draws a line you don't have to defend with willpower. Willpower is also a brain resource. You have less to spend.

Assessment

AI didn't hurt my brain. It also didn't fix it. What it did was close a coding gap built up across three career chapters — from designing rockets, to integrating systems, to running software programs. Each move pulled me further from hands-on building. AI handed that part back. Suddenly I could ship the things I used to only spec, and that's been genuinely life-changing.

But the pace it enables exposed a TBI ceiling I didn't know I had to respect. The fog, the lost context, the sudden crashes — those aren't proof that AI is bad for brain injuries. They're proof that unlimited access to a fast tool, with no guardrails, will push a TBI brain past its limits faster than anything else I've used.

The 2-terminal rule is my answer. It's not perfect. Some days I want to do more. But it keeps me functional, keeps the work sharp, and keeps the fog from owning the day. BCG says healthy workers peak at 3^[9]. TBI research says my resources are reduced^[3]. The math says 2. The experience says the same.

Nobody is studying this intersection yet. AI cognitive load researchers don't look at brain injury populations. TBI researchers don't look at AI tools. Until they meet in the middle, we're running our own trials. Two terminals. Cap before the crash. Respect the threshold.

The barbell gets its Neuro-RPE. The keyboard deserves one too.