TL;DR: Doomprompting is compulsive prompt refinement disguised as work. It creates dopamine loops identical to social media scrolling. Passive acceptance with AI causes skill degradation within 6 months. Active engagement (think, execute, verify) preserves skills. Put the phone down and do the hard thinking.

Saturday morning, I’m on a train with a newspaper and a book next to me - I specifically didn’t bring my laptop so I wouldn’t work this weekend. But I’m already working, on my phone, deep in a chat with an AI.

I don’t even remember the moment I switched, one second I was planning to read, the next I was “just checking” how my prompt training is progressing - no transition, no conscious choice, just that familiar flow.

For a few minutes I felt good about it: “Look at me, I’m being productive, getting ahead on next week”. Then I realized I wasn’t actually working, I was just scratching an itch.

In the past I could lose hours to games or mindless scrolling, and I knew exactly what that was: entertainment, a simple dopamine loop of Input, Action, Reward. Sitting there on the train I realized I hadn’t actually grown out of that loop, I had just swapped the interface: traded the game controller for a chat window, traded the “Level Up” screen for a generated strategy, traded Doomscrolling for Doomprompting.

The Perfect Camouflage

This new loop is a lot more dangerous than social media ever was. When you play a game for five hours straight you feel the guilt, you know you “wasted” time, but when you spend five hours prompting an AI to refine a document - tweaking instructions, regenerating output, asking for one more variation - you feel virtuous, telling yourself you’re being thorough, thinking you’re optimizing.

But are you? Or are you just chasing the “Slot Machine” effect?

The mechanism is exactly like Instagram: you pull the lever (write a prompt), wait those few seconds of anticipation, andding- the reward appears, a perfectly formatted response, a clever solution, a block of text. It’s the same immediate hit as a “Like” notification, but since the output looks like “work” the internal alarm bells never go off. We’ve gamified our own jobs and we didn’t even notice.

The Cost of Passive Acceptance

This isn’t just a hunch - the data is starting to show what this does to our brains. A recent study from Anthropic (January 2026) tracked how people learn with AI assistance and found a huge difference between two specific behaviors.

The first is Active Engagement: you think through the problem first, use AI to execute, then verify the result - this actually works. The second is Passive Acceptance: you let the AI lead, accept the output because it looks professional, let the tool do the thinking.

The researchers found that Passive Acceptance leads to skill degradation: people felt faster but were actually learning less, and six months later they couldn’t do the same task without the AI. This is exactly what “Doomprompting” looks like.

When I’m mindlessly prompting on the train I’m not doing Deep Work, not wrestling with the hard stuff - I’m just letting the AI drive because it feels good to see the text appear. I’m feeling productive (high velocity) but I’m actually stalling (low value).

The Productivity Paradox

This explains the weird disconnect we’re seeing in tech right now: most managers report feeling 20-24% faster with AI, but objective measurements - like the METR study from late January - show that for complex work we are often slower.

Why? Because we’re spending all our time managing the tool instead of solving the problem, debugging prompts instead of debugging logic, reviewing five variations of a strategy instead of just picking one and executing.

Turns out we’ve replaced the difficult, quiet work of thinking with the fast, loud, exciting work of generating.

Reclaiming the Quiet

I’m not going to stop using AI - I use it every day, it’s the most powerful tool I’ve ever had. But I’m realizing I have to treat it with the same caution I use for social media: it’s a high-dopamine environment specifically designed to keep you engaged.

If I can’t sit on a train for an hour with just my own thoughts, without reaching for the “Slot Machine” to generate something, that’s not productivity - that’s a dependency.

So I finally put the phone away and picked up the book. It was slower, a lot quieter, no flashing cursors or instant answers. And for the first time in a week I actually started thinking.

What about you?

Have you caught yourself “Doomprompting”? Do you ever spend an hour refining prompts for something you could have done yourself in twenty minutes?

Let’s be honest about it.

Frequently Asked Questions

What is Doomprompting and how does it work?

Doomprompting is compulsive prompt refinement that masquerades as productivity. You write a prompt, wait for output, refine it, regenerate, ask for variations—chasing the dopamine hit of generated text appearing on screen. It feels virtuous because the output looks like work. It’s not. It’s scratching an itch.

How does passive acceptance with AI cause skill degradation?

Anthropic’s January 2026 study tracked two behaviors: Active Engagement (think first, use AI for execution, verify output) and Passive Acceptance (let AI lead, accept output without critical review). Six months later, workers in the passive group couldn’t do the same task without AI. Passive acceptance trades short-term velocity for long-term skill loss.

Why do managers report faster productivity but complex work is actually slower?

Managers measure velocity (text generated per hour). Objective studies (METR, late January) measure value (problems solved, decisions made). Doomprompting increases velocity while reducing value. Regenerating five strategy variations instead of picking one and executing looks productive but wastes time in refinement loops instead of implementation.

What’s the difference between active and passive AI engagement?

Active engagement: think through the problem first, then use AI to execute faster, then verify the result. This preserves skills and delivers real value. Passive acceptance: let AI lead the thinking, accept output because it looks professional, skip verification. This degrades skills within months and creates false velocity without value.