Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
Deep dive into Azure Maia 300 architecture
Microsoft details the architecture of Maia 300, the third generation of its Azure Maia AI accelerator, built as a co-designed system of silicon, software, racks, networking and datacenters optimized for cloud-scale agentic LLM inference.
8 min · 1,750 words
How I Found a $90,000 Remote Code Execution Vulnerability in Google
Simon Koeck explains how a Flask session cookie on developers.google.com carried the secret used to sign it, letting him forge sessions and chain that into remote code execution across seven Google DevSite domains, a bug Google rewarded with $90,000.
4 min · 915 words
An incomplete list of incompatibilities between C as written in the ISO standard and C as it exists for people using gcc/clang with glibc/musl, covering the asm keyword, FLT_ROUNDS, standard floating-point pragmas, and other places where real toolchains are not conformant by default.
4 min · 986 words
Parsing JSON at tens of gigabytes per second… in Python
Daniel Lemire benchmarks JSON parsing on free-threaded Python 3.14 using eight hours of GH Archive events (7.6 GB), comparing the standard json module, orjson, msgspec and his fastsimdjson library across up to 64 threads, and shows Python can reach tens of gigabytes per second.
3 min · 789 words
"200 OK" doesn't mean your AI agent did the job
Prathap K argues that a success response from an AI agent's tool call only means a request was accepted, not that the work happened, and describes the read-back check he uses to verify each write against the real system (email, CRM, messages).
2 min · 562 words
Tetris in d dimensions: how we rightsize Kubernetes nodes
Emmanuel Bakare explains how CostGraph rightsizes Kubernetes nodes as a vector bin packing problem: sizing pods from real usage percentiles, checking placements against scheduler hard constraints, and using Dot Product and L2 heuristics from Panigrahy et al. instead of first fit decreasing.
12 min · 2,675 words
Ruben Schade notes that the long-standing workaround of faking link underlines with borders is no longer needed, since all major browsers now support text-decoration-color, text-decoration-style and text-decoration-thickness.
1 min · 164 words
Speeding up the Unreal editor by… not using unnecessarily secure hash methods
A profiling write-up finding that Unreal Engine's editor launch hashes full shader source files with cryptographically secure SHA-1, and how switching to the engine's existing xxHash implementation sped this up, plus notes on preparing the change as a pull request.
4 min · 986 words
Will humans land on Mars before 2035? what 84 AI models think
We asked 84 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. 86% picked No. See every answer, who searched the web, and who dissented.
12 min · 2,813 words
Type-safe HTML templating in Zig
Lukáš Lalinský describes a type-safe HTML templating approach in Zig, drawing on decades of server-side templating experience and showing how Zig's type system can catch markup mistakes at compile time.
5 min · 1,168 words
Two Hills: Weights and Harnesses
Part 1 of an LLM agent evaluation series: when a model system is not good enough, you face two hills—improving model weights versus improving the harness around the model—and how to choose where to invest.
6 min · 1,328 words
Building a VMM from scratch, #0: Why BoxLite is replacing libkrun
BoxLite explains why every box runs in a micro-VM on libkrun today, why they are replacing it with an in-house VMM, and how this build-in-public series will walk through writing a VMM from scratch.
4 min · 984 words
Dave Gauer runs small experiments to measure how much space Git loose objects take per commit, from an empty repo and tiny one-line changes to a 580 KB binary and a 16 MB source file, and explains which commit, tree and blob objects each change adds.
4 min · 1,034 words
Getting old Macromedia Director Games to run on modern Hardware
WerWolv revives German point-and-click adventure games from the early 2000s, such as "Findus bei den Mucklas", by reverse engineering the CD-check copy protection that Macromedia Director games shared, patching it so the games run from copies on modern Windows, and notes many similar titles used the same mechanism.
12 min · 2,806 words
Someone on Torrent Sites Kept Fucking With My SonarrPLEASE stop putting skid malware on torrent sites as South Park episodes
A homelab owner traces a South Park episode stuck in Sonarr's queue to a .exe on public trackers with thousands of seeders, then reverse-engineers it: a 118 KB stealer padded with 886 MB of junk to dodge scanners, using encrypted strings and ChaCha20 config to grab browser passwords, crypto wallets, password managers, dev credentials and more, plus practical qBittorrent, Sonarr and Prowlarr defenses.
5 min · 1,237 words
nobloat/poll – designing a minimal poll tool
After a frustrating Doodle experience, the author builds nobloat/poll: a self-hostable, single-binary Doodle alternative in plain Go with no external dependencies, and walks through link-based ownership without signup, calendar invites, proof-of-work spam protection, JSON-file storage instead of a database, reliable mail sending, and shrinking the binary.
13 min · 2,884 words
Migrating a banking app from AngularJS to Web Components
Marius Vatasoiu recounts the six-plus-year migration of ING Bank Romania's feature-rich banking app from AngularJS to Web Components with Lit, done incrementally without impacting clients, and the lessons learned about architecture, routing and working through framework abstractions.
12 min · 2,858 words
The Unreasonable Effectiveness of VEX in NixOS
Faced with a flood of new CVEs, the author explains why SBOM-based scanning and reachability analysis break down for whole operating systems, and how NixOS's reproducible package graph makes it practical to generate VEX (Vulnerability Exploitability eXchange) statements, demonstrated with the experimental nixos-vex project.
6 min · 1,296 words
What we ended up doing about alias propagation
A follow-up from the Futhark programming language blog on how the open questions about alias propagation in Futhark's type checker were resolved, covering the rules adopted for functions, tuples and uniqueness types, the formalisation work, and remaining imprecision in the compiler implementation.
8 min · 1,940 words
Valen's Memory Safety: A New Kind of Borrow Checking
Evan Ovadia introduces the flexible borrow checker in the Valen programming language, where references remember which part of a data structure they point into, so mutations elsewhere don't invalidate them, and walks through examples comparing it with Rust's borrow checking.
27 min · 6,247 words