Beginner again — with 25 years of IT behind me

The Sunday edition: a longer read for the second coffee. After passwords and the AI-ify reflex, something personal today: a story about starting over.

There is a sentence that sits verbatim in my work history, typed somewhere between two configuration files: «I still have to learn Omarchy!» With an exclamation mark. You can read the mild panic right off it.

For context: I have worked in IT for more than 25 years. I have been responsible for a company’s worldwide IT — data centres, migrations, budgets. And then I sat in front of my new laptop with a deliberately chosen, opinionated Arch Linux system — and wanted to edit a file. No vi installed. Fine, the other editor then. Once inside nvim, I failed at copying cleanly. Copying. The operation I have performed daily since the last millennium.

That same week — physically in a deck chair, the holidays had officially started — I asked beginner questions a first-semester student would have been embarrassed by: What is tmux actually for? How do passkeys work with the fingerprint sensor on Linux? Why does this window not do what I want?

This edition is about why that was the best working week in a long time.

The fear nobody says out loud

Let us talk about the feeling first, because it is not exclusive to IT people. «Starting over» is one of the most expensive fears in working life. It keeps developers on dying technologies, companies on systems nobody wants to touch anymore, and whole careers inside the comfort zone. The maths behind it was always rational: relearning cost weeks and months — manuals, forums, trial and error, and nobody you could ask something embarrassing at three in the afternoon without losing time or face.

In companies the same fear is called «our people can’t do that». It is the secret reason the old system is still running even though it annoys everyone: it is not the replacement that scares people, it is the afterwards — the time when everyone is a beginner again.

What was different this time

This time a tutor sat next to the beginner. Not a course, not a video, not a forum — an AI agent with three properties no human teacher and no documentation ever had at the same time:

It knows my system, not some system. The answer to «why does this window not do what I want?» was in no manual, because it depended on my configuration — which the agent knows, because it co-wrote it. Every answer arrives in the context of my machine, not as a generic recipe that first needs translating.

Questions cost nothing. No waiting, no appointment slot, and — this is psychologically bigger than it sounds — no shame. The tmux question from the deck chair was the third «stupid» question that day. With humans, you budget which questions you can afford. Here you simply ask. Anyone who remembers how much they learned as a child by asking non-stop understands what that means.

The answer stays. Whatever comes up twice becomes a runbook in my setup — written down by the agent itself. The classic learning loss («I solved this once before…») disappears, because the reference grows with every solved problem. The third time, it is routine.

The result in numbers, honestly estimated: switching to a system that would have cost me a month of friction in the old world was productive within days. Not because I got smarter — but because a learning curve is no longer steep when someone stands at every step and explains.

What the tutor does not take off your hands

Honesty is part of the format. Three things even the best AI tutor does not do for you:

• The muscle memory. Understanding how to copy in the editor is one thing; doing it automatically and correctly at eleven at night is another. Practice remains practice.
• The judgement. The agent explains three ways and recommends one. Whether the recommendation fits my priorities is mine to assess — and for that I need to understand enough. Whoever adopts blindly is not learning, but delegating the thinking.
• The start. No tool makes the decision to step back onto the lowest rung. It is uncomfortable even with a tutor. Just no longer expensive.

What this means for companies

Now the transfer, because it is the actual point. When the most expensive fear in working life suddenly gets cheaper, calculations shift that spoke against change for years:

«Our people can’t operate the new system» was a heavy argument in 2020 — trainings, productivity loss, frustration. In 2026, the same patient tutor that sat next to me can sit next to every employee: one that knows the company’s specific system, answers every question in context, and turns the answers into a growing, company-owned documentation. The precondition is the same as always in this series: the tutor must be allowed to see the system and write the docs — documented processes, accessible systems. AI-ready, in other words.

The punchline: of all the things I have AI-ified in this rebuild, learning is the most underrated one. Everyone talks about passwords and reflexes. Hardly anyone talks about the learning curve — the oldest argument against every change — going flat right now. Yet that is exactly what changes the maths.

The Sunday question

Which change have you — personally or as a company — postponed lately because the relearning seemed too expensive? And would that calculation still come out the same today, with a tutor next to every chair? If not: maybe a deck chair is not the worst place to become a beginner again.

Questions, beginner stories of your own? Write to me — I answer personally. See you next Sunday.