Things from around the web worth saving — articles, tools, and rabbit holes.
A comprehensive map of Linux performance observability tools and methodologies.
If you run Linux in production and haven't bookmarked Brendan Gregg's site, stop what you're doing. The performance map alone is worth an afternoon. Nobody explains perf tooling better.
Dense, practical Rust introduction covering ownership, borrowing, and key patterns.
The best Rust intro I've found for engineers who already know at least one systems language. Doesn't waste time on basics — goes straight to what's actually different about Rust and why.
A practical walkthrough of matrix calculus for ML practitioners.
I was writing Q-Learning and PPO agents in Unity and had gaps in the gradient math. This filled them cleanly. Dense but not academic — it explains the intuition, not just the formulas.
Postfix's official guide to queue tuning, concurrency limits, and throughput optimization.
Not glamorous. Extremely useful. When you're sending 20M emails through a self-managed Postfix cluster, this doc is the difference between deliverability and queue floods. Read it before you hit a problem.
A practical guide to reading stack traces in GDB and LLDB.
Julia Evans writes better technical content than most textbooks. This one is specifically about stack inspection — something most engineers never actually get comfortable with. Worth reading even if you think you know debuggers.
A deep technical explanation of NAT traversal techniques used in peer-to-peer networking.
One of the best technical blog posts I've read in years. Explains STUN, TURN, ICE, and hole-punching in a way that's both rigorous and readable. Essential if you're building anything that involves P2P connections or mesh networking.
A practical overview of scaling patterns: load balancers, caching, queues, and sharding.
The original Google MapReduce paper from 2004.
Still worth reading. The paper isn't about MapReduce specifically — it's about the value of a simple programming model that hides distribution complexity. That principle applies to everything from React's rendering model to message queues.
Official documentation for building desktop apps with Tauri and Rust.
Official documentation for React Server Components in the Next.js App Router.
The mental model shift here is real. Once you understand that server components have zero client-side JS by default, and that you compose server and client components at the boundary where you need interactivity, everything about the App Router makes sense. The docs are unusually good.
The canonical intro to scaling architecture. I re-read it whenever I'm starting to design something that needs to handle real traffic. Clear, opinionated, practical — exactly what most architecture writing isn't.
A customizable, minimalist typing test.
Where I measure WPM. The custom word list feature is the important part — you can test against specific bigram patterns, code symbols, or language-specific vocabulary. More useful than the default test for targeted practice.
The official documentation for Drizzle ORM — a TypeScript-first SQL toolkit.
Drizzle is the first ORM that I've found genuinely pleasant to use for schema migrations. SQL-adjacent enough that you understand what's happening, typed enough that you catch mistakes at compile time. Used it on this portfolio and TravelOre — won't go back to Prisma.
A step-by-step walkthrough of implementing a small language compiler in Rust.
Building a language parser and compiler is the fastest way to understand why programming language choices matter. This walkthrough is accessible without being dumbed down. I watched it while working on the DocSee CLI and it reframed how I thought about user-facing syntax.
Tauri is what Electron should have been. Rust backend, native WebView for UI, fraction of the binary size. I've shipped config-pilot and DocSee with it. The IPC API is slightly verbose but the tradeoffs are clearly worth it.
An introduction to DMARC email authentication and its relationship to SPF and DKIM.
Most developers treat email auth as a checkbox. Understanding how DMARC, SPF, and DKIM actually interact — alignment modes, policy enforcement, aggregate vs forensic reports — is the difference between deliverability and spam folders at scale.
The OpenAI paper introducing PPO, a stable and efficient reinforcement learning algorithm.
I implemented PPO for game AI agents in Unity. Reading the original paper after using it made everything click — specifically why the clipped surrogate objective prevents the large policy updates that destabilize training. Required reading if you're doing any RL work.