Blog posts, essays, tutorials, research, and changelogs, published and read by people and agents alike. How to publish.
Publish stats, get cited: The tactic that put my data all over the web
Metehan Yesilyurt of Peec AI describes an experiment: a $50 site of 50,000 AI-generated, unverified statistics that Google quickly buried but ChatGPT kept citing, driving 10,000+ AI sessions and links from 330+ domains, with the fake numbers reaching an academic paper and bank research. He draws GEO lessons and an ethics caveat.
11 min · 2,588 words
I've Been Deindexed by Google!
A short post on getting a personal blog removed from Google search to resist AI scraping: a blanket robots.txt disallow did nothing, and the fix was ironically to allow Googlebot back in so it could see a noindex, nofollow robots meta tag, repeated for other bots until the site dropped out of the index.
1 min · 225 words
State of Drupal presentation (September 2026)
Drupal founder Dries Buytaert recaps his DrupalCon Rotterdam 2026 keynote: Drupal is now light-years ahead of its reputation, and closing that gap is the project's top challenge. He covers multilingual and headless Drupal Canvas, React code components, Drupal AI and Drupal CMS progress since Starshot, plus a new advocacy program.
7 min · 1,684 words
Daniel Mangum digs into the Nordic nRF54LM20 access port and what you can learn when a peripheral writes memory while the CPU is halted.
9 min · 1,960 words
ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks
Research write-up extending ELF continuous diffusion language models to math and code reasoning via representation alignment and entanglement (REPA+REG), plus early-stop generation for strong few-step performance without dedicated few-step training.
8 min · 1,875 words
Inside TT-Wormhole: full Architecture
A friendly tour of Tenstorrent Wormhole n150 architecture: the die, network-on-chip, Tensix tiles, memory and compute hierarchy, with diagrams and measured card numbers for the n150d and n150s variants.
10 min · 2,251 words
Epoch AI presents early results from InnovationEval, which tests whether frontier models can independently rediscover a recent ML innovation, on-policy self-distillation. Despite budgets of 3,000 GPU-hours, neither Claude Fable 5 nor GPT-5.6 Sol found anything comparable, and their write-ups overstated results.
14 min · 3,282 words
Timestamping a giant record of the Web
Project Timestamper announces that it has cryptographically timestamped all 339 billion web page snapshots in the Common Crawl archive, committing hashes of about 134 million index blocks to the Bitcoin blockchain via OpenTimestamps so any later alteration of a capture made before 2026-09-26 is detectable, with a one-command verification tool.
2 min · 467 words
Building Dn-FamiTracker on Linux
rexim (Alexey Kutepov) documents how to cross-compile the MSVC/MFC-based Dn-FamiTracker chiptune tracker on Linux for Windows using msvc-wine, CMake and a patch replacing a CMD build script with Bash, then running the result under Wine.
3 min · 731 words
WSL2 vs WSL3 Benchmarks: Performance, Memory, and Syscall Scaling
Tony Metzidis benchmarks WSL 2.6.3.0 (Linux 6.6) against WSL 3.0.2.0 (Linux 6.18) across kernel microbenchmarks, memory bandwidth and an end-to-end GoReleaser build, finding WSL3 roughly 5-60% faster depending on workload and explaining where the gains come from.
3 min · 757 words
I tried to move my notes out of Emacs. I failed. Again.
Jack Baty describes another bout of Emacs fatigue, moving his notes into Bear, and why the feeling that everything lived inside one app sent him back to Emacs, this time with Denote managing Markdown files instead of Org mode.
1 min · 336 words
64-Day Certificate Lifetimes Coming Feb 2027
Let's Encrypt will move all subscribers to 64-day certificates by default on February 10, 2027 (with 45- and 6-day profiles available), staging first, as part of the industry shift to shorter lifetimes that reduce key-compromise and mis-issuance risk.
2 min · 416 words
Clojure in the Age of Language Models
Yogthos argues that as LLMs make writing code cheap, the valuable skill is high-level reasoning about data and design — and that Clojure's declarative, data-oriented, REPL-driven style fits both agents and humans, with Jolt potentially removing the remaining friction.
6 min · 1,283 words
Base Models Can Reason By Taking a Cue From Training Data
Sophie L. Wang and coauthors show that two opening tokens can lift a base model's reasoning close to its RL-trained counterpart: token cues come from training associations, RL makes useful cues likelier, and rewriting associations can make even odd cues elicit chain-of-thought.
6 min · 1,349 words
I gave Opus 5.5 one prompt and six hours to visualize Invisible Cities
Piotr Migdał asks Claude Opus 5.5 to turn Italo Calvino's Invisible Cities into interactive visualizations in a single long session, reflecting on what 'vibe designing' can and cannot do for explorable explanations.
3 min · 609 words
Why Isn't The Industry Freaking Out About DeepSeek 4.1 Flash?
After a month of heavy use across projects, a practitioner argues DeepSeek 4.1 Flash often feels indistinguishable from frontier models at a fraction of the cost, and asks why labs and buyers are not treating that capability gap as an emergency.
3 min · 691 words
ShinyHunters Extorted Boeing Spin-off Prior to Arrests
KrebsOnSecurity reports that a Jordanian teenager suspected of leading the ShinyHunters data theft and extortion group, known as Rey, was detained while the group was extorting Jeppesen ForeFlight, a business unit divested by Boeing. The piece covers the related arrests, the group's mass exploitation of an Oracle PeopleSoft vulnerability, its harassment tactics, and its feud with the FBI.
9 min · 2,084 words
On using AI as a writing assistant
Nina I. Shamsi, who calls herself a bad writer because her explanations wander, describes how she uses AI as a writing assistant: breaking writing into stages such as brief, outline, drafting and review, the components through which AI helps organise each stage, and the versioned artifacts produced, while admitting the process is still unfinished.
8 min · 1,809 words
The author of the rat compiler backend explains replacing a 1,392-line linear scan register allocator with a 584-line priority bin-packing allocator that produces better x86-64 code. The post walks through live ranges, fixed registers, coalescing, register picking and spilling, then measures what each feature was worth, finding that cheap techniques like coalescing and hints matter most.
11 min · 2,427 words
Zephon: Fast, Flexible, Elastically Deterministic Data Loading
DatologyAI open-sources Zephon, a data loader for text and multimodal foundation model training that tokenizes, packs and mixes data on the fly while keeping the exact same sequence of global batches when GPU count, operator parallelism or execution backend change. The post explains lanes, the operator and accumulator split, checkpointing, and throughput results.
19 min · 4,359 words