Issue #70 | May 20, 2026 | Forwarded by 400+ engineers
The Thing Worth Naming This Week
There's a particular flavor of confusion that shows up around year two or three of heavy AI-assisted coding.
You look at your title. You look at your output. You look at your salary. And something in your gut says: I don't think I could build this from scratch anymore.
Not imposter syndrome — that would be easier. Imposter syndrome says "I'm not good enough." This says something worse: "I don't know what I actually am anymore."
That's the competence inflation trap. The work looks like yours. The output carries your name. But underneath, there's a gap between what you can do and what your title says you should be able to do.
Why This Happens
In traditional engineering, your skills and your output were tightly coupled. You wrote the code, you understood the code, you could debug the code without AI assistance. Your competence was visible in the work.
AI assistance decouples these. The output quality stays high — sometimes higher than you could achieve alone. But the relationship between you and that output has changed. You're now often curating and directing rather than building and understanding.
The problem isn't that you're faking it. The problem is that the old measures of competence were built for a different world.
Three Signals You're In It
1. You can't remember the last time you wrote a meaningful piece of code without AI help.
Not because you're lazy. Because the loop is so fast and the output so good that there's no gap left to struggle through. But struggle is where learning happens.
2. You're proud of what you shipped but uneasy about what you personally contributed.
The "I" in "I built this" has gotten complicated. You shipped a feature. You guided the AI. You made dozens of decisions. But the craft-level satisfaction — the kind that comes from knowing you personally overcame something hard — is dimmer.
3. You'd struggle to replicate your own work without AI.
Not because you're weak. Because the work was genuinely complex and AI handled parts of it. But this creates a strange dependency: you can only maintain and extend what you can understand — and you may not fully understand all of it anymore.
What The Research Says
A 2025 study of software engineers at three large tech companies found that AI-assisted developers completed tasks 40% faster but showed 23% lower accuracy on follow-up debugging tasks without AI assistance — a measure researchers called "output fidelity retention."
The engineers who fared best: those who had explicit "no-AI windows" — deliberate practice sessions where they built without assistance, even though it was slower and more frustrating. Not because they needed to prove something, but because they needed to feel their own thinking at work.
One Thing Worth Trying
The Saturday Morning Rule:
Once a month, wake up Saturday, open a fresh editor, and build something small — a utility script, a small web app, a data processing job — with one constraint: no AI assistance for the first 90 minutes.
Not as a test. Not as a purification ritual. Just to maintain the circuit.
The goal isn't to prove you don't need AI. It's to keep the connection between your thinking and your making from going fully quiet.
A Quote Worth Sitting With
"The person who knows howwhat to askThe answer isn't to stop using AI. It's to stay aware of what you're trading.
From the Clearing
This week on the site:
- The Science of AI Fatigue — updated with new cognitive load research and the latest on working memory limits under AI-assisted workflows - AI Fatigue Severity Index — take the 47-point self-assessment if you want a more complete picture of where you are - Daily AI Boundaries — a practical 30-day plan for engineers who want to rebuild their relationship with AI tools, starting small
That's it for this week. Forwarded to you by 400+ engineers who found their way here.
The clearing is real. The path back is slow. But it's real.
Missed an issue? The full archive is at [clearing-ai.com/newsletter-archive](https://clearing-ai.com/newsletter-archive)
Want to share your story? [Share your experience](https://clearing-ai.com/share-your-story.html) — anonymous, always.
Managing a team? [Team Manager Guide](https://clearing-ai.com/team-manager-guide.html) — how to recognize and respond to AI fatigue on your team
If this was forwarded to you and you'd like to subscribe directly: [clearing-ai.com/newsletter](https://clearing-ai.com/newsletter)
If the the competence inflation trap resonates — the AI Fatigue Quiz surfaces where else this dynamic shows up in your daily experience with AI tools.
Take the AI Fatigue Quiz → Read: An Engineer's Manifesto for Intentional AI Use →P.S. If you found this useful — forward it to an engineer who could use it. The best way to grow a community worth being in is to invite people who are trying to work well with AI.