Featured post what's this? ✨ AI is destroying Open Source, and it's not even good yet Over the weekend Ars Technica retracted an article because the AI a writer used hallucinated quotes from an open source library maintainer. The irony here is the maintainer in question, Scott Shambaugh, was harassed by someone's AI agent over not merging ...
Introducing Sonnet 4.6 Claude Sonnet 4.6 is a full upgrade of the model’s skills across coding, computer use, long-reasoning, agent planning, knowledge work, and design.
GrapheneOS - break free from Google and Apple [ENG 🇬🇧] 🇬🇧->🇵🇱 Przejdź do polskiej wersji tego wpisu / Go to polish version of this post
Rise of the Triforce During the rapid technological advancements of the early 1990s, the video game industry was on the cusp of a massive addition - another dimension. With console shenanigans like the Super FX chip giving players a taste of 3D, hype was at an all-time high. ...
LLM-generated skills work, if you generate them afterwards LLM “skills” are a short explanatory prompt for a particular task, typically bundled with helper scripts. A recent paper showed that while skills are useful to LLMs, LLM-authored skills are not. From the abstract:
SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks Agent Skills are structured packages of procedural knowledge that augment LLM agents at inference time. Despite rapid adoption, there is no standard way to measure whether they actually help. We present SkillsBench, a benchmark of 86 tasks across 11 domai...
A/B Testing Your RAG Pipeline: Chunking, Retrieval, and Reranking Strategies You Can Build With One Prompt Each | Rasha Hantash How to quickly build and compare RAG pipeline variants — cosine vs. hybrid search, fixed vs. semantic chunking, Cohere vs. cross-encoder reranking — using Claude Code, Graphite stacks, and your own offline evals.
Plan it, Work it, Review it, Reflect it Our workflow for working with AI agents: create, plan, implement, review and reflect.
Rob Panico - You Never See the Whole Shape at Once A tesseract can never be seen all at once—only through partial, time-bound projections—and that limitation turns out to be the point. What looks contradictory from a single frame often reveals coherence when allowed to rotate over time. This essay uses th...
Experiments with CodeMirror - Blog - aziis98 This is a tutorial on writing a simple code review interface with CodeMirror 6 and the unified merge extension.
Evaluate Your Own RAG: Why Best Practices Failed Us - Charles AZAM We benchmarked our production RAG system across embedding models, chunk sizes, chunking strategies, and retrieval modes. The results contradicted common wisdom.
On Dyslexia, Programming and Lisp. — Relections on Software Engineering Dyslexia is a difference in the way the brain processes language. It is often defined as the difference between intelligence (vaguely defined via the fake IQ test) and a person's ability to read. That is, a dyslexic with normal intelligence will underperf...
Things I Check Before Opening a PR What is a programmer but a series of PRs (pull requests)? I optimize PRs to introduce the best code I can, be easy to review, and document my work so I can make sense of it in the future. Here are some things I always check before opening a PR.
Personal Software | Matt Spear I built a habit tracker, a tab manager, a task board, and an AI assistant. None are products. They're personal software – and soon everyone will have their own.
Deploying Your Own IndieWeb Site with Indiekit + Eleventy (Docker Compose based) Deploying Your Own IndieWeb Site with Indiekit + Eleventy (Docker Compose based) 14 February 2026 indiekit indieweb deploy eleventy A complete guide to deploying [Indiekit](https://getindiekit.com) on...
The Long Tail of LLM-Assisted Decompilation After rapid advances thanks to one-shot decompilation, progress on the Snowboard Kids 2 decompilation began to falter. This post explores the workflow evolution, tooling improvements, and fundamental LLM limits that emerged when tackling the long tail of ...
Brent Benson - A custom app creation renaissance There is a change in the calculus of what is achievable with a sketch and a 10 minute conversation
A Love Letter to Self-Hosting I’ve tried to put to paper why I love self-hosting; yet I’ve never been fully able to accurately describe what I love about it. Perhaps it is the digital sovereignty, or the rejection of living on the internet and renting my space, or perhaps it is a way ...
The speed of building has outpaced the thinking: why we need a new moral standard for AI development Explore the impact of AI on indie development and the need for a moral compass in coding. Are we sacrificing quality for speed?
Harnessing Postgres race conditions Synchronization barriers let you test for race conditions with confidence.
Do LLMs hallucinate more in Czech than in English? – Miloš Švaňa Having explored hallucination benchmarks for LLMs, I’ve decided to use the TruthfulQA dataset to see if LLMs hallucinate more when I talk to them in Czech instead of English.
I built a coding agent two months before ChatGPT existed I built a coding agent back in 2022, 2 months before ChatGPT launched:
Why do I feel bad following recommendation algorithms? How can I still find interesting things without them?
Peer-reviewers in the coal mine? About AI & employment Maybe Section 174 killed Brynjolfsson canaries. Maybe it's indeed AI. This article is about how you should treat working papers as working papers and give more love to the process of peer-reviewing.
Words are a Leaky Abstraction Something about hearing the phrase "Claude's Soul Document" got me thinking again about a problem I've long pondered. Software folks, like myself, have a long and proud history of taking words that exist and coopting them, manipulating the meaning of thos...
Get a portable monitor, they’re great An exciting new form of gadget — portable monitors! Light, 10” to 21”, USB-C power, signal via USB-C or mini-HDMI. And portable.
Use Protocols, Not Services The Internet is almost anonymous and privacy-preserving by design. I mean, unless some administrator actively tries to track you, there is no built-in...
Khronos Announces glTF Gaussian Splatting Extension Today, The Khronos Group, announces a release candidate for the…
AI is destroying Open Source, and it's not even good yet Over the weekend Ars Technica retracted an article because the AI a writer used hallucinated quotes from an open source library maintainer. The irony here is the maintainer in question, Scott Shambaugh, was harassed by someone's AI agent over not merging ...
