When AI Helps You Go Fast—But You Lose Your Mind

How generative AI is creating a new kind of debt that lives in our heads, not our code.

Remember when “moving fast and breaking things” was Facebook’s motto? We used to worry about technical debt—messy code, hasty architectures, shortcuts that would haunt future developers. But here’s the uncomfortable truth: AI might have solved technical debt while introducing something far more insidious.

The Core Insight

Margaret-Anne Storey coined a term that hit me like a freight train: cognitive debt. It’s the idea that the debt accumulated from “going fast” doesn’t live in your codebase—it lives in your brain. Even when AI agents produce code that looks clean and functional, the humans involved may have completely lost the plot. They can’t explain why certain decisions were made, how parts of the system work together, or worse—how to change anything without breaking everything.

Here’s the anecdote that sealed it for me: A student team she coached hit a wall around week 7-8. They couldn’t make simple changes without breaking something unexpected. At first, they blamed “technical debt”—messy code, poor architecture. But dig deeper, and the real problem emerged: no one on the team could explain why certain design decisions had been made or how different parts of the system were supposed to work together. The code might have been messy, but the bigger issue was that their shared understanding had fragmented entirely.

Why This Matters

I’ve experienced this myself on my more ambitious “vibe-coding” projects. I’ve prompted entire features into existence without reviewing their implementations—and while it works surprisingly well, I’ve found myself getting lost in my own projects. I no longer have a firm mental model of what they can do and how they work. Each additional feature becomes harder to reason about, eventually reaching a point where I lose the ability to make confident decisions about where to go next.

This is the paradox of AI-assisted development: we’re more productive than ever, but our understanding of what we’re building is shallower than ever. The code exists, it works, but we can’t explain it—and that gap between output and understanding is cognitive debt.

Key Takeaways

• Technical debt lives in code; cognitive debt lives in heads — Same outcome (inability to move fast), completely different root cause
• AI makes it worse before it makes it better — You can ship faster, but you understand less
• The cure isn’t less AI—it’s more intentional oversight — Regular code review, even of AI-generated code, is critical
• Teams are more vulnerable than individuals — When multiple people rely on AI to build, shared understanding erodes faster

Looking Ahead

The future of software development isn’t about choosing between AI assistance and human understanding—it’s about finding the right balance. The most successful teams won’t be the ones shipping fastest with AI; they’ll be the ones who figured out how to maintain cognitive coherence while leveraging AI’s power.

Maybe it’s time we added “maintaining shared understanding” to Definition of Done. Because shipping features you can’t explain isn’t progress—it’s just faster accumulation of debt.

Based on analysis of Margaret-Anne Storey’s “Cognitive Debt” via Simon Willison