Everything on the outside looks fine.
Your velocity is up. Your output is consistent. Your PRs are on time. Your sprint estimates are accurate — or at least, the tickets are getting closed. Your performance review from last quarter was solid. Nobody has flagged anything wrong.
And yet: something inside has gone quiet.
Not burned out. Not depressed. Not lazy. Something quieter than that. A hollowness that does not have a name in the language the industry gives you. You have been performing normalcy so successfully that even you sometimes forget something is wrong.
This is the outside-in problem.
The industry measures what is measurable. Story points closed. PRs merged. Incidents resolved. Features shipped. These metrics are visible. They appear in dashboards. They inform performance reviews. They determine who gets promoted and who gets marked as a concern.
What the metrics cannot see:
These things do not appear in any dashboard. They accumulate invisibly. And because they are invisible, they get dismissed — by you, as much as by anyone else. It can't be that bad. My metrics look fine.
The cruelest version of this is when engineers start to gaslight themselves. When the external signals say everything is fine but the internal signal says something has gone wrong — the internal signal is almost always right.
The outside-in problem is that the industry gave you a set of metrics to optimize, and those metrics are now so far from what actually matters that you are winning by every external measure while losing something you cannot name.
You might recognize some of these:
You can close tickets but not answer questions about them. You ship features. You cannot, without the AI in front of you, explain why the approach was chosen. The code works. You could not have generated it at the speed you needed it without help. And you have started to notice the gap and not say anything.
Your satisfaction metrics are wrong. You do not feel good about shipped features. You feel relief that it is over. Relief that the ticket is closed. The difference matters enormously — one is the end of building, the other is the end of performing.
You have stopped raising edge cases. Not because you do not see them. Because you do not have the energy to both see them and also prompt your way through explaining why they matter. You let things go. The code works. It will probably be fine.
You can no longer tell when you are actually thinking. You prompt. The AI generates. You evaluate. You ship. Somewhere in that sequence there used to be a moment of genuine deliberation — a pause where you weighed options, considered trade-offs, made a call you could defend. That moment is gone now. The sequence is so fast you cannot find the gap where thinking used to happen.
Imposter syndrome says: I am not as capable as people think.
The outside-in problem says something different: I am performing at a level that is visible, while the thing that used to make the work meaningful has become inaccessible.
These are not the same. Imposter syndrome is a perception problem. The outside-in problem is a functional one. You can close tickets at high velocity and still be experiencing the outside-in problem. The performance and the loss can exist simultaneously — and they do, for most engineers who are navigating this.
The starting point is the least comfortable one: you have to stop relying on the external metrics to tell you the truth about how you are doing.
The metrics will tell you everything is fine until it is very not fine.
So you need a different check. Something personal. Something that does not show up in any dashboard. Something like:
These checks are uncomfortable. They reveal the gap before the gap becomes a crisis. They are also the first step toward doing something about it — because you cannot rebuild something you have not first acknowledged is missing.
Every week, thousands of engineers take the AI Fatigue Quiz on the site. The data keeps showing the same pattern: the outside-in problem is not rare. It is the majority experience.
The numbers are large because the problem is common. You are not weak for noticing it. You are paying attention — which is the first step toward doing something about it.
Where are you on the outside-in spectrum?
The quiz tells you your specific profile and what the gap between your output and your understanding actually means for your situation. Takes 90 seconds.
Friday afternoon. Pick one ticket from the sprint. Do not open any AI tools until you have spent 20 minutes thinking about it first. Open the file. Read the context. Write down your hypothesis before you prompt for anything.
Then prompt, build, ship. Compare your hypothesis to what the AI generated. If the comparison surprises you — even if the AI is better — the surprise itself is information. It tells you something about the gap between your model and the tool's model.
That gap is where the thinking used to happen. And it is still there. You just have not been using it.
See you next Tuesday.
— Sunny + The Clearing team