Why this exists

A few years ago I was sitting in a design review, nodding along while someone explained why we needed Kafka. I understood the words. I did not understand the thing. And I’d been writing software for over a decade at that point.

I don’t think I’m unusual. Most of us collect a working vocabulary — Kafka, Raft, B-trees, eventual consistency, backpressure — without ever building the mental model underneath. We can use these things. We can wire them up and ship. But we can’t quite explain them, and it shows the moment someone asks the only question that matters: “but why?”

The 2am version of this problem

I’ve spent a lot of nights on-call. And the incidents that actually scared me were never the ones with a clean error message. They were the ones where I realised, somewhere around the third hour, that I didn’t really understand the system I was paging about. I was pattern-matching. Hoping. (Two things you do not want to be doing at 2am.)

So I started going back and properly learning the fundamentals I’d been hand-waving past for years. Turns out the stuff underneath is more interesting than the framework-of-the-month on top of it. It’s also the stuff that doesn’t go stale.

And then there’s the AI thing

There’s now a model that’ll write the code for you. Which is genuinely useful — I use it every day. But it quietly changes the math: it makes the shallow stuff cheap, and the deep stuff valuable. When anyone can generate a working implementation, the person who actually understands the system — why it works, where it breaks, what to reach for — is the one who’s still in the room. Fundamentals are the moat. So that’s where I want to spend my time.

So this is the fix I’m building

One concept a week, explained the way I wish someone had explained it to me. The problem it solves first. Then a mental model you can actually hold in your head. Then the real mechanics. Then the gotchas nobody warns you about. No listicles, no “in today’s fast-paced world”, no 800-word warm-up before the point. Just the thing, understood properly, in about five minutes.

I’ll be honest about how it’s made, because I’d want that as a reader: every issue is generated with AI, then fact-checked — by a second model and by a human — before it goes out. I’m not going to pretend I hand-write each one at 2am. (I tried. It doesn’t scale, and you’d get about four of them.) But nothing publishes without a person saying “yes, this is correct.” If something’s still wrong, there’s a report button, and it gets pulled back for review. I’d much rather be corrected than confidently wrong — that’s the whole point of the site.

Who’s behind it

I’m Sachin — an SRE and platform engineer who’s spent more years than I’ll admit keeping other people’s systems alive. I’m also building 2AM Agent, an open, self-hostable agent that investigates incidents so the on-call human doesn’t have to start from a blank terminal at the worst possible hour. This site and that project share one belief: understand the system, and the 2am pages get a lot less scary.

If that resonates, the weekly’s free. One email, every Week, no noise.