Why I'm starting to write this down

Early on at the new job, someone asked me why adding a retry to a particular call had made a failure worse rather than better.

I had fixed that exact failure before. Twice, in a previous codebase. I opened my mouth to explain it and found that what I had was a memory of the incident rather than a reason. I knew the shape and the thing that worked. I could not say why it worked, which meant I could not tell that person whether it applied to what they were looking at.

Two boxes: in my head, I already know this, versus in a sentence, where does it actually start. — Figure 1. The comfortable version lives in your head. The real one has to survive a sentence.

Four years of learning by incident

I spent four years at Menlo Security, joining as an intern and leaving as a Senior Engineer. I learned the craft inside a single codebase: deeply, and narrowly. You absorb an enormous amount that way. You also pick up assumptions you cannot see, because you have never worked anywhere that made different ones.

Then I moved to Acceldata. Different pace, different domain, and a set of habits I did not know I was carrying until they stopped being true.

One codebase, deep and narrow, then a new team where the old assumptions show. — Figure 2. The move did not create the gap. It made it visible.

Production teaches by incident. You learn networking because a tunnel misbehaved. You learn observability because something broke and nobody could see why. The knowledge is real. It is shaped like a list of things that have already happened to you, not a set of principles you can point at a problem you have not met yet.

Changing teams did not create that gap. It put me in rooms where I had to say things out loud. The gap was not in what I had done. It was in what I could explain.

Two boxes: used it, from incidents in one codebase, and can explain it, a model you can reuse. — Figure 3. The gap is not skill. It is a model you can say out loud.

Writing is the fix I settled on, for a boring reason. It is much harder to fool yourself in a paragraph than in your own head. A sentence has to commit to something. Once it is written down you can notice that it does not actually follow.

What this is going to be

Notes on systems, networking, and observability. Mental models from data structures and algorithms: the patterns, not the problem numbers. Write-ups of things I have actually hit. Occasionally the parts of this job that are not technical.

What it will not be is a set of polished takes with the conclusion picked in advance. If a post changes its mind partway through, that is the process showing. I would rather publish something honest and be corrected than publish something safe and learn nothing from it.