The writers whose archives are worth reading end to end: Ciechanowski's interactive essays, Dan Luu, rachelbythebay, Bert Hubert, Fabien Sanglard, and the rest of the highest signal-per-post writers on the internet.

The freely available math that punches far above its fame: Evan Chen's Napkin, MIT's Mathematics for CS, Linear Algebra Done Wrong, and the Boneh–Shoup cryptography book with cryptopals as its gym.

Where the GPU-performance knowledge actually lives: PMPP, GPU MODE, Simon Boehm's matmul worklog, Horace He's brrrr essay, the Ultra-Scale Playbook, and Stas Bekman's open ML-engineering book.

The language-implementation canon that actually gets read: Crafting Interpreters, Russ Cox's essays, Eli Bendersky's archive, matklad, and the Architecture of Open Source Applications.

The database-internals material practitioners actually learn from: Use The Index Luke, CMU 15-445/721, Database Internals, the Red Book, and the blogs where people build storage engines in public.

The under-known heavyweights of low-level systems: McKenney's perfbook, Drepper's memory paper, Agner Fog's manuals, Bakhvalov's free performance book, 1024cores, Preshing, and the linker essays.

The CP resources serious competitors use that rarely make beginner lists: Steven Halim's CP4 and Methods to Solve, VisuAlgo, Sannemo's free book, KACTL, and the Codeforces Catalog.

Where to actually learn probability and statistics: Stat 110, Wasserman's All of Statistics, and the applied habits — intervals, bootstraps, and paired tests — that every honest benchmark depends on.

The distributed-systems path: Kleppmann's DDIA as the map, MIT 6.824 labs as the practice, Jepsen as the reality check, and the five papers worth reading in full.

The performance canon: Brendan Gregg's book and methods, napkin math, latency numbers, and the discipline of benchmarking honestly.