The 30-second version of a skill's half-life

Three things have a half-life: uranium-238, the enthusiasm of a new hire, and your knowledge of CSS.

The first is on a poster in a physics classroom. The second is on a Slack channel everyone politely doesn't post in anymore. The third is the largest of the three, and almost nobody is willing to say it out loud, including, especially, the people whose paychecks depend on knowing CSS.

This post is the thirty-second version of an idea that takes about fifteen minutes to fully unpack. I think the thirty-second version is enough to change how you spend tomorrow morning, which is mostly what I'm hoping for.

Here is the idea

The half-life of a skill is the time it takes for half of its useful application to be gone — not gone from your memory, gone from the world. The skill of typing has a half-life measured in decades. The skill of writing AngularJS 1.x had one that ended on a specific Tuesday in 2016.

The trap most working people fall into is mistaking high-skill for long-half-life. They are not the same thing. There are people in our industry who are spectacularly skilled at things that will be irrelevant within five years, and there are people in our industry who are quietly mediocre at things that will still pay rent in 2050. The second group — the ones with the boring, durable skills — are the ones whose careers don't terrify them at three a.m.

The good news is that the long-half-life skills are not secret. They are mostly things like: knowing what a good question feels like; reading code like prose; smelling a bad abstraction; knowing when to stop. None of these will appear in a job listing. All of them are what the job listing is actually trying to find. They are also — and this is the inconvenient part — mostly invisible until you've done the work for ten years.

The bad news is that nobody is going to teach them to you. Not your bootcamp. Not your university. Not your company's training budget. The institutions that could teach the long-half-life skills are run by people who need to pay rent next month, and their incentives are aligned with selling you whatever's hot this quarter.

What to do tomorrow morning

Three things, all small, all free, all of which compound:

Read code older than you are. The Linux kernel. SQLite. Anything written in C in the eighties that's still running. The patterns that show up across decades of unrelated codebases are the patterns that reflect properties of the problem, not properties of the framework. Those are the patterns with the long half-life.

Write postmortems for things you didn't break. Pick a public outage — Cloudflare, Knight Capital, S3 — and write your own postmortem from scratch. The exercise trains the part of your brain that recognizes what was missing upstream of the failure. You will start seeing the same gap in your own work within weeks.

Pick one quarter of your reading time and aim it twenty years out. Database internals. Operating systems. Type theory. The bones. It will not feel productive. It is precisely what makes you productive a decade from now.

That's the thirty-second version. There's a longer one, behind the paywall, that gets into the math, the diagnostic, and the part nobody tells you — about why the people who actually have these skills are also, almost universally, kinder than the people who don't.

It is one of the strangest correlations in the field, and I think it might be the most important.

→ The full essay: the half-life of a skill.

If you like this kind of writing, the full version is part of the Late-Night Curriculum — a paid series of essays that build on each other. One sharp idea per post. One signature diagram per post. No corners cut. The kind of thing you'd send to a friend at 1 a.m. without apology.