---
title: "A mentor that never says no"
in_short: "Answers got easy, and understanding didn't. Lean too hard on any automation and the skill behind it quietly fades. Use new tools in a way that still makes you think."
canonical: https://www.hontolab.com/en/blog/a-mentor-that-never-says-no
language: en
authors: "Katarina Kostic"
published: 2026-11-03T09:00:00Z
modified: 2026-11-03T09:00:00Z
publisher: Hontolab
---

# A mentor that never says no

> **In short:** Answers got easy, and understanding didn't. Lean too hard on any automation and the skill behind it quietly fades. Use new tools in a way that still makes you think.

You have an answer. Do you know the answer?

Knowing what to ask, and whether the answer holds, is still the hard part. Picture learning before the internet: no search bar, no chat, just a library card and whoever you could ask. A mentor pointed you in the right direction, and then you learned by struggling to understand. That struggle kept the ability to learn, and the persistence behind it, in shape.

Search engines took away most of the hunt, and made it easy to skip the mentor too, although a good one still beats the first page of results. You still had to read and form the insight yourself. Now even that looks optional: the answer arrives fully formed, exactly as you asked it, which is a problem when you asked the wrong thing. Before AI, you had to understand the problem to get anywhere. With AI you still do. It just no longer makes you.

## Warm water

A good mentor gives you a nudge and lets you work it out. Chat tools do the opposite; as one recent study puts it, AI almost never says no. It is instant and genuinely helpful in the moment, and like the boiling frog, you don't notice what it does to your thinking until the water gets warm.

## Assistant, rather than substitute

You have probably heard it from the people around you, or experienced it first hand: using AI as part of daily work can feel like a superpower at first, then slowly bring less patience for the hard part and more work handed over without being understood.

The study behind the mentor line, [AI Assistance Reduces Persistence and Hurts Independent Performance](https://arxiv.org/pdf/2604.04721), put that feeling to the test. People solved school-level maths and reading problems with an AI chat beside them, until it was taken away without warning, less than a quarter of an hour in. Without it, they did worse on average than people who never had it: they had the answers, but they didn't know them.

After the chat was taken away, it came down to how people had used it. Those who used it like a spotter, asking for hints, did about as well as people who never had the chat. Those who let it lift for them, asking for final answers, did worse and gave up more often. That matters: persistence is one of the strongest predictors of long-term learning. A pattern rather than a proof, but a clear one: an assistant supports your thinking, a substitute replaces it.

Nobody has yet measured what years of this might do. The concern is that tools designed to make things easier in the moment may come at the expense of your ability to learn and improve over time.

## Keeping yourself in the loop

The researchers doubt that less time with AI, or a Socratic AI that answers with questions, would be enough. Still, it is a good place to start. What helps is reorganising the interaction loop: let the model question you, give hints and leave room to try first.

Imagine a mentor you can ask anything, any time: late at night when a question comes up, or on a weekend when a lightbulb goes on about your project. A mentor like that is only as good as the questions you bring to it, and “What day is it?” is not one of them.

## Your idea, then the drafts

"A human had the idea. AI ran the drafts. The human decided which one held." – the common sign-off on Hontolab blog posts covers more than the posts themselves. It is how we try to work with AI in general. Running the drafts is the start of the conversation, and the decision stays with you. Notice when you reach for the chat before you have an idea of your own, or when thinking without it feels harder than it used to. That is the water getting warmer.

Two fair objections: if you can always have the answer, why bother knowing it? And at work, it is often not up to you: handing things over to AI is becoming part of the job. Neither changes the cost: over time, careless use can wear down the persistence it takes to stay with the next hard problem.

Having the right answer seems easy now, but only as long as you know what to ask and can tell whether it holds. That part was always your job; nothing makes you do it anymore.

**Katarina Kostic**

*The thinking was human. The commas were negotiable.*