top of page

AI Did the Work. Your Self-Trust Didn't Grow.

Writer: Steffen Moessner
Steffen Moessner
Oct 25
5 min read

Updated: 5 days ago

The short version


  • AI can finish the task. It cannot hand you the proof that you could have.

  • Self-trust is built from evidence, and the evidence comes from hard things you got through yourself.

  • If someone always stepped in for you early, this lands on an old nerve.


I keep hearing a version of the same sentence this year. Someone walks me through a strong week. Work shipped, a hard email finally sent, a problem solved by Friday. Then it arrives, quieter: "I don't know if I could have done any of that on my own."


That isn't really a complaint about AI. It's someone noticing they've lost track of what they can do.


A climber partway up a rock face, deciding whether to keep going

What Self-Trust Is Made Of


Self-trust isn't a mood you talk yourself into. It's a record. It gets built the slow way, out of moments where you didn't know how something would go, stayed with it anyway, and came out the other side. The not-knowing is what the record gets made of.


Which is why the most useful part of a hard problem is rarely the answer. It's the twenty minutes before the answer, when you were stuck and had to decide whether to keep going.


Anthropic ran a small study on exactly this. Fifty-two developers, mostly junior, learned an unfamiliar Python library. Half had an AI assistant, half didn't. Afterwards everyone took a quiz with no AI allowed. The assisted group averaged 50%. The unassisted group averaged 67%. The widest gap was in debugging, which is the part where you sit with something broken and don't yet know why.


The detail I keep coming back to: it wasn't whether people used AI. It was how. The ones who asked it why something worked kept more than the ones who handed over the hard part.


Before we go on, which wire is it in you? I built this quiz around the early imprints I work with in coaching. Seven questions, under two minutes, and it names the one running strongest right now. Where it comes from, and what to do with it, is the work we do together.



Borrowed Competence Feels Like Your Own


There's a second problem, and it's quieter. Researchers at Aalto University gave people logical reasoning problems, some with ChatGPT and some without. 246 people, then 452 more in a replication. AI raised their scores. It also raised how well they thought they'd done, past what they'd actually earned. The people who knew the most about AI judged themselves least accurately.


Worth saying plainly: the group without AI overestimated themselves too. People do that. What AI changes is that nothing in the loop corrects it, and being good with AI doesn't protect you from that.


In practice that means the evidence you use to answer "can I do this?" gets contaminated. The work went well. Your name is on it. The feeling of having been capable arrives right on schedule. Only the part that builds self-trust is missing, and nothing tells you it's missing.


Why It Lands Harder on Some People


Plenty of people lean on AI all day and feel fine about it. The ones who end up across the table from me are usually the ones where it caught an old wire.


If you grew up with someone who stepped in every time things got hard, who finished the homework, made the phone call, smoothed it over before you had to, you learned something specific. Not that you're incapable. That you're capable as long as someone is there.


My own coach, Ramón Schlemmbach, gives each of these imprints a sentence you can catch yourself saying. For this one it's: "I can't figure this out on my own." Almost nobody says it out loud. They just notice they've opened the chat before they tried.


This isn't about bad parents. Usually the help came from love and simply arrived a little before it was needed. The cost shows up decades later, in someone who has never collected much proof of their own.


Then a tool arrives that is endlessly patient, never irritated, awake at two in the morning, and never once suggests you try it yourself first.


It Happens to Experts Too


There's an uncomfortable study out of four Polish hospitals. After AI tools were introduced for colonoscopies, the doctors' detection rate when working without the AI fell from 28.4% to 22.4%, across 1,443 procedures.


I want to be careful with that one. It covered 19 doctors, it watched what happened rather than testing it properly, and other researchers have pointed out real holes. Treat it as an early warning, not a verdict. But it's worth sitting with, because these were people who already had the skill before the tool showed up.


What Helps


I'm not going to tell you to stop using it. That advice gets ignored, and it should be. What matters is where the struggle still happens.


  1. Try first, then ask. Ten minutes of your own attempt before you open the chat, even a bad attempt. The attempt is the part that leaves a mark.

  2. Ask why, not do. "Why does this work" builds something. "Write this for me" doesn't. That was the difference inside the same tool in the Anthropic study.

  3. Predict before you check. Write your own answer down first, then compare. You get honest feedback on your judgment instead of a borrowed result.

  4. Keep one unassisted zone. One thing a week you do the hard way on purpose.


Self-trust updates the way any old belief updates, by collecting evidence against it. You can't argue yourself into believing you're capable. You can only keep putting things on the pile that say otherwise, and one hard thing finished unassisted is worth more than a hundred reassurances.


Questions people ask me about this


So should I stop using AI?

No, and you wouldn't anyway. The question is which parts you've stopped doing yourself, and whether you chose that or drifted into it.


How do I know if I've actually lost something?

Try one thing cold, with no help, and notice what happens in your body before you start. Reluctance is normal. Dread is information.


Isn't this the same panic people had about calculators?

Partly, and that's fair. A calculator took one narrow step off your hands. This takes the judgment, and judgment is what self-trust is made of.


My job requires it. Now what?

Then the unassisted zone lives outside work. An instrument, a language, repairing something. The subject matters less than getting back the experience of being stuck and staying.


Where does childhood actually come into it?

In how fast you reach. Someone who was left to work things out hesitates before asking. Someone who was rescued reaches immediately, and has done it so long it doesn't feel like a choice.


If This Is Landing


Reaching for help isn't weakness and it never was. It became a reflex because at some point it was the thing that worked. What's worth rebuilding is the part of you that knows what happens when nobody comes, because that part is where the confidence you keep looking for lives.


Sources: Anthropic (2026), "How AI assistance impacts the formation of coding skills". Fernandes et al., "AI makes you smarter but none the wiser", Computers in Human Behavior. Budzyń et al. (2025), endoscopist deskilling after exposure to AI in colonoscopy, The Lancet Gastroenterology & Hepatology.



Keep reading


Steffen Moessner is a life coach in Palo Alto who works online with clients worldwide. He helps people find the patterns they learned early and change how they react today. Trained at the Co-Active Training Institute.

bottom of page