When AI Limits Become Superpowers: Breaking Mental Roadblocks in Neurodivergent Development

The Invisible Productivity Killers

Mental roadblocks used to kill my productivity. I’d have a brilliant idea, then spend three hours staring at an empty Jira epic template, paralyzed by “proper Agile formatting.” By the time I figured out the corporate language, I’d lost the creative thread entirely.

Sound familiar? If you’re neurodivergent—particularly dealing with ADHD like I am—you know these invisible barriers all too well. Certain things would just stall my brain and slow or halt my productivity on work that actually mattered. When I work on my own projects, I like to just go. I don’t like to do a lot of planning. But I’ve also learned that planning is good, and the more complex things get, the more you need some sort of dynamic, living documentation to fall back on.

The cruel irony? The very tools designed to help manage complex projects often become the biggest obstacles for neurodivergent brains.

The Planning Paradox

Here’s the tension every ADHD developer faces: We thrive on creative flow and rapid iteration, but complex projects genuinely need structure and documentation. Traditional project management tools force us into rigid formats that feel like creativity killers.

In my previous articles, I shared how AI transformed my development workflow—first with explosive productivity gains, then through hard-learned lessons about limitations and discipline. But what I didn’t fully appreciate until recently was how AI could solve an entirely different problem: translating between how my ADHD brain naturally thinks and the rigid structures that corporate tools demand.

The Breakthrough: AI as Cognitive Bridge

Then something revolutionary happened. I discovered that Claude AI integrates seamlessly with Jira for project management. But it wasn’t just the integration—it was how it worked.

Now I just tell Claude what I’m thinking, and it handles the translation. This isn’t just productivity—it’s removing invisible barriers that neurodivergent brains face every day.

One of my biggest roadblocks with tools like Jira was my ADHD brain hitting walls around format and proper terminology. I’d waste enormous amounts of time trying to craft the “right” language for Agile epics and stories. Now I can talk to Claude however my brain thinks, and it understands and updates Jira for me. The cognitive load just… disappeared.

Here’s what a typical interaction looks like now:

The initial prompt:

AI automatically generating appropriate Jira Epics:

Adding an Epic to a Story:

Successfully completing a story and having it updated in Jira:

The transformation is immediate:

• Before: ADHD brain → forced into Jira’s rigid format → cognitive exhaustion → procrastination → project stalls
• Now: ADHD brain → natural conversation with Claude → perfectly formatted Jira tickets → momentum maintained

When AI Limits Forced Better Habits

But here’s where the story gets really interesting. As I continued building The Shadow Cortex (TSC)—my ambitious security platform designed to help with ADHD roadblocks when faced with large amounts of data and choices—I started hitting AI usage limits.

Google Gemini claims to handle millions of tokens for context. Claude honestly limits everything to 200k. And you know what? That limit turned out to be helpful.

I’m past my twenties, and I’ve learned over the years how to adapt to deal with my ADHD. This was another situation where I had to leverage my adaptability and find another way—a workaround. And you know what? It made everything better.

The Enforced Testing Revolution

Those context limitations forced me into something I’d been avoiding: rigorous end-to-end testing. I realized I could work within AI’s constraints by creating a feedback loop:

1. AI writes code (within context limits)
2. AI creates unit tests for that code
3. AI runs the tests
4. AI fixes any failed tests
5. Repeat until 100% passing
6. I, the human, manually verifies and tests everything afterwards

This AI-enforced TDD workflow completely eliminated the infinite loop problems I encountered with mixed Svelte 4/5 syntax in my previous articles. Everything I’m building into the new version of TSC has been rock-solid.

Building Assistive Technology with AI

The meta-narrative here is fascinating: I’m using AI to build an AI-assisted system (TSC) that helps my ADHD brain handle complexity. It’s assistive technology built with assistive technology.

TSC is becoming exactly what I need: a tool that presents complex security data in ways that don’t overwhelm my neurodivergent brain. Clean interfaces, logical groupings, smart filtering—all designed to work with how my mind processes information, not against it.

The Testing Infrastructure as Cognitive Support

What I didn’t expect was how comprehensive testing would become another form of cognitive scaffolding. When you have ADHD, complex systems can become overwhelming quickly. But with solid test coverage, I can:

• Make changes confidently without fear of breaking everything
• Understand system behavior through test descriptions
• Catch regressions immediately before they compound
• Maintain focus on new features instead of debugging old ones

I’m now integrating:

• 50+ passing Go unit tests for backend logic
• Frontend component tests for UI reliability
• Playwright end-to-end tests for complete user workflows
• Gosec for Go security scanning
• OWASP ZAP integration (soon) for automated security testing

Each layer provides a safety net that lets my creative brain run free while ensuring nothing breaks.

The Beautiful Irony of AI Limitations

Here’s what I find most amusing about this journey:

• Google Gemini: Claims massive context windows but loses track of conversations
• Claude: Honest about 200k limits, which actually helps you work better
• “Lesser” AI models: Force better practices that produce superior results

Sometimes the best AI feature is knowing when to stop helping you.

The constraints didn’t limit my productivity—they channeled it. By forcing me to work in smaller, testable chunks, AI limitations actually made my code more reliable and my process more sustainable.

Key Takeaways for Neurodivergent Developers

Embrace AI as Cognitive Translation:

• Use AI to bridge between your natural thought patterns and corporate tool requirements
• Don’t force yourself into rigid formats—let AI handle the translation
• Focus your mental energy on creative problem-solving, not bureaucratic formatting

Turn Limitations into Disciplines:

• AI context limits can enforce better coding practices
• Constraints often lead to more robust, testable code
• Sometimes “worse” tools force better habits

Build Assistive Workflows:

• Design your development process around your neurodivergent strengths
• Use testing as cognitive scaffolding, not just code validation
• Create safety nets that let you experiment without fear

Leverage Your Adaptability:

• ADHD brains are exceptionally good at finding creative workarounds
• Experience + adaptability + constraints = breakthrough innovation
• What feels like a limitation might be pointing toward a better approach

Looking Forward: AI That Works With Neurodivergent Minds

My work on TSC continues, but now with a proven methodology that works with my ADHD brain instead of against it. The combination of AI-assisted planning, natural language project management, and enforced testing discipline has created a sustainable development workflow.

More importantly, I’m building something that might help other neurodivergent professionals handle information overload and decision paralysis. If AI can help translate my scattered thoughts into organized Jira epics, imagine what purpose-built assistive technology could do.

The future isn’t just about AI making us more productive—it’s about AI making complex work accessible to minds that think differently. And sometimes, the best innovations come from working around limitations rather than trying to eliminate them.

Building tools for neurodivergent minds? I’d love to connect and share experiences. Find me on LinkedIn or check out my ongoing work at j0n-harr150n.com.