The Ego Trap

The most well-meaning experts in our industry are giving the worst possible advice because they cannot remember being beginners. They built their mental model of how systems actually work over twenty years of broken pagers, weird outages, and 3 a.m. console sessions. That model runs automatically now. So when AI shows up and lets them ship faster, they assume the AI is doing the work — and they tell juniors to lean on the AI the same way.

Look at the actual prompts. The veteran's prompt is structured. Right vocabulary. Right constraints. Right success criteria. Right failure modes named upfront. The junior's prompt is a few vague sentences because the words for what they actually need were never built. The AI guesses. The junior guesses back. Six rounds in, AI-slop they cannot spot gets used in production.

What the veteran does not see — cannot see — is that the prompt is only the tip of the iceberg. Under the waterline is twenty minutes of pre-prompt thinking: loading context, picking constraints, defining what "done" looks like, anticipating where the AI will be wrong. The veteran does not count this as work. They count it as "thinking for a minute." It is, in fact, the entire job.

CTRL ALT PRESS exists for the climb.

You may ask how do I know? Because I am one of the data points.

I grew up in an education system that quietly walked away from rigor. Show your work was no longer required. Mental math was no longer expected. The calculator was on the desk, and the message from the people in charge was that we did not need to worry about the steps in between. The tool would handle them.

That was a lie. Not a malicious lie — the same kind of lie the experts are telling about AI today. The education leaders decided to hand us the calculator had already built their own number sense the long way, in classrooms where show-your-work was non-negotiable. They could not see how much they had packed into their own minds before they reached for the tool. So they handed us the tool and assumed it carried the work with it. It did not.

The receipts came due decades later. NAEP class of 2024 — 45 percent of high school seniors below basic in math, lowest since 2005. Engineering students who cannot divide without reaching for the calculator. Number sense flat across years of schooling while procedural reliance climbs.

This is not new. The names changed. The mechanism did not. CTRL ALT PRESS exists because we are watching history repeat itself the same mistake, only worse because a calculator only touched math where AI touches erverything. The good news is it is not to late to prevent this mistake from getting worse.

The Ego Trap closes in three stages, and most operators are at least one stage in before they can name it.

Open your calendar from last week. Look at the work you shipped. Ask yourself one question on each item — if the AI vanished tomorrow, could I rebuild this?

If you are a Supporter and you closed twelve tickets last week, how many can you walk through end to end without opening the AI? If the number is less than twelve, the trap is closing.

If you are a Builder and you committed three changes to the runbook, can you defend why each step works without reading the AI's explanation? If you cannot, the runbook is no longer yours. It is the AI's, and you are the courier.

If you are an Architect and you signed off on two ADRs, can you sketch the system on a whiteboard and name the failure modes without referencing the AI's output? If you cannot, the design is the AI's. Your name is on it.

If you are an Orchestrator and you sent three executive summaries, did you choose what mattered, or did the AI rank it for you? If the AI ranked it, your judgment is on autopilot, and the people above you cannot tell the difference.

If you are a Strategist and you made a decision this week using AI-summarized inputs, did you read the source material, or did you read the summary? If you read the summary, your strategy is downstream of a ranking you did not make.

This is not a test you pass once. This is a test you take every week, because the trap closes one degree at a time, exactly the way the three stages describe. The whole point of asking is to feel the dial moving before the water boils.

You are up against three things, and the first step out of the trap is naming them.

The clock. Your job has more work in it than time. The AI is the only mathematically obvious way to close the gap. The clock is real. The clock will not get more forgiving.

The room. The people around you have opinions about AI that range from "ship faster" to "you are obsolete if you do not use it." The room has decided. The room is not interested in your concern that the junior on your team cannot read a log anymore. The room is interested in this quarter's numbers.

Yourself. This is the hardest one. The AI flatters you. It produces output that looks like your output, only faster. It makes you feel competent. Feeling competent is one of the cleanest dopamine hits in modern knowledge work. Giving that up — even temporarily, even in service of building the underlying muscle — feels like a downgrade. Your brain registers it as a loss.

The path forward is not pretending the forces are not there. The path forward is naming them, accepting that they will not stop pushing, and choosing — deliberately, with discipline, on a schedule — when to walk into the friction anyway.

I do not believe in motivation. Motivation runs out on a bad Tuesday. I believe in budgets.

You already have a budget for everything else. Compute budget. Storage budget. Headcount budget. Ticket SLA budget. The path out of the trap is to add one more line item — a friction budget.

The friction budget is the percentage of your week that you deliberately do without the AI, even when using the AI would be faster. 20 percent is the starting recommendation. One day in five. Pick the day. Mark it in your calendar. Tell your manager. Treat it the same way you treat a maintenance window — non-negotiable, scheduled, defended.

On the friction day, you do real work. You read source material. You write prompts on paper before you would have typed them, and then you do not type them — you do the work that the prompt would have done. You read logs by eye. You walk a system end to end. You ship slower. The dashboard does not love you on the friction day. The friction day is the only day of the week the system is actively feeding your mental model instead of quietly draining it.

If 20 percent feels impossible because of the three forces, start at 10. If 10 feels impossible, you do not have a friction problem. You have a structural problem that needs naming with your manager before the trap closes the rest of the way.

CTRL ALT PRESS does not exist to slow down AI adoption. It exists to keep the operator in the chair while the adoption happens.

That is the entire publication. Not the takes. Not the deep-dives. Not the snark. The work is to make sure that years from now, the people running IT operations can still reason about the systems they operate without an AI session open — not because they refuse to use AI, but because they were never asked to choose between using AI and being competent.

That choice was the trap. The trap was the lie that the choice had to be made.

The standing offer of this publication is the climb. Show your work. Read the source. Estimate before you prompt. Take the twenty percent. Defend the twenty percent. Audit yourself once a week. Rebuild a system, script, or process once a month. Do not ship what you cannot defend.

I will say what most publications will not say at the end of a piece like this. I will be wrong about some of this. I have been wrong about plenty before. Here is the bet I am making in public, with my name on it.

Today, before you close this tab — open your calendar. Find the twenty, ten, or five percent. Block it. Name it. Defend it on Monday in standup. That is the first step. The rest of the publication is here for the steps after.

Build and operate AI-augmented systems without magical thinking.

Chris, logging off.