Neurodivergent Engineers & AI Fatigue:
A Different Kind of Overwhelm

🧠 ADHD: The Dopamine Loop Problem

ADHD brains are driven by novelty-seeking and dopamine dysregulation. AI coding tools are, neurochemically speaking, a perfect storm:

• Every AI suggestion is a micro-novelty that triggers a dopamine hit — compounding an already hyperactive novelty-seeking system
• AI enables rapid context-switching, which ADHD brains find more stimulating than sustained focus (but which destroys deep work)
• The passive acceptance of AI suggestions bypasses the effortful engagement that actually builds competence and confidence
• Rejection Sensitive Dysphoria (RSD) — common in ADHD — turns AI code rejection into an emotional crisis, making engineers over-rely on AI to avoid perceived disapproval

The inattentive subtype faces the highest risk: passive AI use looks identical to productive engagement from the outside, but compounds attention deficits invisibly.

🔵 Autism: The Processing Load Problem

Autistic engineers often have heightened sensory sensitivity, deep systematic thinking, and higher baseline cognitive load from social-navigation at work. AI tools add to that load in specific ways:

• AI outputs are ambiguous by design — partial implementations, "try this", hedged suggestions. For engineers who process through precise, systematic frameworks, this ambiguity is cognitive friction, not helpful flexibility
• Constant AI suggestions add a second processing layer on top of already heightened environmental awareness
• Social demands at work (explanations, code reviews) already consume more cognitive resources for autistic engineers; AI tool adoption adds another domain of "keeping up"
• Monotropic attention (deep focus on specific interests) can make switching to AI-maintained languages/frameworks feel like identity loss

Many autistic engineers have strong systematic thinking that AI can't replicate — but AI's suggestion loop can cause doubt about skills that are genuinely strong.

💔 Rejection Sensitive Dysphoria: The Approval Problem

RSD isn't a formal diagnosis but is extremely common in ADHD and often present in other neurodivergent profiles. It means social-evaluative situations — performance reviews, code critiques, technical disagreements — trigger outsized emotional responses that feel like pain.

AI interacts with RSD in a specific trap: the safest code is AI-generated code. If you've internalized that AI is "smarter" or "better," accepting AI suggestions feels safer than submitting your own work. This is not laziness — it's a neurological hypersensitivity to perceived failure being managed as well as possible.

The compounding effect: over-reliance on AI → reduced skill confidence → more RSD around code quality → more reliance on AI to avoid the pain of rejection. A self-reinforcing loop that feels like relief and looks like good performance, until it doesn't.

🌡️ Sensory Sensitivity: The Overwhelm Problem

Engineers with sensory processing differences (common in autism, ADHD, and other neurodivergent profiles) often have less cognitive bandwidth available after baseline sensory navigation. An office environment, screen brightness, ambient noise — these aren't neutral backgrounds for neurodivergent engineers, they're additional cognitive load.

AI tools add processing demands that are invisible in ergonomic reviews: parsing suggestion UI, evaluating correctness, maintaining a model of what the AI "knows," managing the workflow of accepting/rejecting. For a brain already running at higher baseline cognitive load, these micro-decisions add up fast. The result isn't procrastination or laziness — it's genuine bandwidth exhaustion.

1. Pair AI with Structured Retrieval, Not Passive Acceptance

For ADHD brains: AI works best as a starting point, not an ending point. The specific practice: read an AI solution, then close the AI, and write/explain the solution from memory without looking at the output. This retrieval practice builds the skill AI temporarily provided, rather than allowing it to atrophy.

Why it works for ADHD: Retrieval feels like "real work" and produces dopamine from genuine comprehension — unlike passive AI acceptance, which is novelty-seeking without learning. For autistic engineers, it also reduces ambiguity by forcing a concrete, personal model of the solution.

2. The External Commitment Device for RSD-Driven Over-Reliance

If your AI over-reliance is driven by RSD (fear of code rejection), externalize the validation. Specifically: commit to writing at least one function per PR from scratch — not because AI can't do it, but because the practice of writing code without AI is the skill you're protecting. Track this externally (a simple tally, a note in your journal).

Why it works for RSD: External commitment shifts the evaluation criteria from "is this code good enough?" to "did I do the practice?" The first question invites anxiety; the second is a simple binary. This is borrowed from exposure therapy frameworks — deliberate, low-stakes practice of the feared behavior.

3. Sensory Budget Management for High-Load Days

Track your sensory-cognitive load before adding AI demands. On days when baseline load is high (poor sleep, social demands, a noisy environment), your capacity for AI's ambiguity-processing is lower. On those days, use AI for narrower, more well-defined tasks — not for exploration or learning.

Why it works: This is proactive load management, not reactive guilt. Instead of "I used too much AI today and feel worse," it's "I knew my bandwidth was low, so I used AI strategically." This reduces the secondary shame that compounds AI fatigue for neurodivergent engineers.

4. The No-AI Initiation Block for ADHD Task-Avoidance

ADHD brains struggle with task initiation. If you're not starting because AI isn't available, that's a signal. Practice: when you sit down to code, use AI only after you've made genuine progress on your own — even 10 minutes of unaided work. The initiation gets done with your own cognitive resources; the AI augments from a position of genuine starting momentum.

Why it works: You're not restricting AI use — you're sequencingit so the dopamine-seeking impulse is served after genuine engagement, not instead of it.

5. Systematic Documentation for Autistic Ambiguity Intolerance

When AI produces an ambiguous or partial solution, develop a personal protocol: write down what the code does, what it assumes, what it doesn't handle, and what "success" means for this specific implementation. This is documentation, but it's also disambiguation — converting AI's probabilistic output into your own precise model.

Why it works: This converts AI output from "something to accept or reject" into "something to understand." For autistic engineers who process through systematic frameworks, this makes AI output cognitively usable rather than ambiguously overwhelming. It also creates a retrievable record of what you learned, not just what the model generated.

6. Weekly Deep Dive Sessions for Monotropic Attention

Block 2-3 hours once a week for monotropic deep work in your specialization — no AI, no Slack, no meetings. This isn't about "proving" you can do it without AI; it's about maintaining the depth of engagement that makes your specific skills valuable and that AI can't replicate.

Why it works: Monotropic attention is a feature, not a bug — deep focus on a specific area is where genuine expertise lives. Protecting that time, even once a week, maintains the depth that counteracts AI's horizontal pressure (broader, shallower coverage across more areas).