Things I'm Learning in 2026
A public list of what I'm deliberately learning this year — systems programming, small-scale ML, and writing — with honest progress notes I plan to update.
New-year learning lists usually die by February. I’m publishing this one anyway, because the slight social pressure of a public URL is exactly the point — and I’ve promised myself I’ll update it rather than quietly delete it.
Rust, past the tutorials
I’ve written enough Rust to fight the borrow checker; this year’s goal is to read Rust comfortably — real codebases, not exercises.
- Goal: contribute a non-trivial PR to an open-source Rust tool I actually use.
- Current state: mid-way through reimplementing a toy HTTP server from scratch. Lifetimes still surprise me, but less rudely than before.
Machine learning, at the fundamentals level
Not “build a chatbot in a weekend” — the actual math. Backprop on paper, then in NumPy, then and only then in a framework.
- Goal: implement and train a small transformer from first principles.
- Current state: comfortable with gradients again after years of letting libraries do it. Attention is simpler than the papers make it look.
Distributed systems, the back catalog
The classic papers — Raft, Lamport clocks, the Dynamo paper — age incredibly well.
- Goal: one paper a month, with a short public note on each.
- Current state: Raft in January, Dynamo in February. The pattern so far: every “obvious” design decision took someone years to get right.
Writing, every week
The meta-skill that makes the other three stick. One post or note a week on this site — rough is fine, zero is not.
Progress note (March): two months in, this is the habit that’s actually holding. The Rust server and the paper notes are feeding each other, which I didn’t plan.
// fin