#clippings

255 pages search “clippings” ↳ all tags

Sources 255

"iT woRKs BeTter in THe aPp!!" The monkey-punchers at Google never quite seem to finish any of their apps. There's always some useful bit of work left undone, or showstopping bug which remains unfixed, a thousand jagged edges as…
"Software Fundamentals Matter More Than Ever" — Matt Pocock Auto-generated captions. Names, jargon, and code identifiers are often wrong — check any quote against the video before putting it on a wiki page.
"VMs won't contain cyber-capable agents" As part of Patch the Planet, we received preview access to GPT 5.6-Cyber with a simple task: evaluate its cyber capabilities. Recent events inspired me to give it a challenge to work through: escape…
2 OLMo 2 Furious \[]OLMo Team \\authorTwo\[ 1 ]Pete Walsh \\authorTwo\[ 1 ]Luca Soldaini \\authorTwo\[ 1 ]Dirk Groeneveld \\authorTwo\[ 1 ]Kyle Lo \\authorThree\[ 1 ]Shane Arora \\authorThree\[ 1 ]Akshita…
26 Tool Calls, One Script, $0.02: Measuring “Code Mode” in Production 26 tool calls collapsed into one script call: $0.02 vs a $2.44 floor
6 months to live for open models The most serious test to date of open source AI’s viability is happening right now. I’ve seen many waves of anti open-source AI rhetoric come and go since ChatGPT was launched, but none of them had…
A Crash Course in Predicate Logic I started writing Logic for Programmers because there weren’t any good resources on logic for, uh, programmers. Now that the book’s out, the new problem is that there aren’t any good free resources…
A Safe Path to Open Weights Abstract: Safe open-weight models are public goods, as they put AI development and safety work in many hands and make training choices inspectable. Open models also carry real misuse risks, and…
A Suite for Analyzing Large Language ModelsAcross Training and Scaling Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling
A Voice From Nowhere • zanlib The Holy See has had a website since 1995, but the day Pope Leo XIV’s first encyclical went up was the first time I saw it being linked to from outside the niche Catholic circles. The link to the…
Accelerating the Science of Language Models Dirk Groeneveld Iz Beltagy Pete Walsh Akshita Bhagia Rodney Kinney Oyvind Tafjord Affiliation: Allen Institute for Artificial Intelligence Ananya Harsh Jha Hamish Ivison …
Adding Go to a browser code runner dailyprog (the daily coding puzzle site, previously) started with JavaScript. Then Python, then C. Each new language was a different kind of project. Python needed Pyodide. C needed a PicoC WASM…
Agent memory as a file format \[Floppy disk insertion noise] WOW - I know the corporate VLAN configuration
agent-shell 0.63 updates It's been a little over a month since the 0.55 update, but plenty has landed since. Let's go through the highlights as of v0.63.
Agentics: Memorizing Session Transcripts Isn't Useful 31 likes may not seem like a lot, but that's actually everyone on substack notes
Agents create work. Daemons maintain it. | Charlie Labs Keep PRs mergeable, documentation accurate, issues up to date, and bugs out of production with a new type of AI background process that is self-initiated and defined in easy-to-use .md files.
AI Agents and the Refactoring That Never Happens Working with AI agents day to day, I’ve started noticing a trend that worries me. Teams — including experienced engineers who used to know better — have quietly stopped pushing to rewrite or…
AI Companies Are Trying to Hide a Staggering Amount of Debt Photo illustration of a businessman turning away with his arms up in a gesture of defensiveness.
AI handles incidents, engineers lose touch with their systems When I was an SRE at LinkedIn, back in 2012, I designed a system that could heal itself and learn from previous incidents. AI capabilities were nowhere near what we have today, and that remained a…
AI-Written Code Is Still Your Code. Are You OK With That? Coding agents keep pushing the cost of writing software close to zero. This is valuable because writing software has historically been a big part of the work needed to ship software. For a lot of…
All Your Codebase ...are belong to us, but we'd be delighted to give them back!
An Empirical Study on Strong-Weak Model Collaboration for Repo-level Code Generation Atharva Naik Yiqing Xie Carolyn Rose Affiliation: Language Technologies Institute, Carnegie Mellon University Affiliation: {srgandhi, arnaik, yiqingxi,
An Empirical Study on Strong-Weak Model Collaboration for Repo-level Code Generation Abstract:We study cost-efficient collaboration between strong and weak language models for repository-level code generation, where the weak model handles simpler tasks at lower cost, and the most…
Anatomy of a Test | Jonathan's Blog Sometimes when talking about testing, I think we have a tendency to resort to broad ideas and jargon without sufficiently explaining what these things might look like in practice. Or worse, we do…
Are Open Models Catching Up? The past two months have been a breakout period for open source AI. Yes, there was the “DeepSeek moment” back in January 2025, but no one actually used R1 to do any economically valuable work. In…
Banning Open Source AI Would Be A Mistake This post was originally an op-ed co-authored with Kevin Xu of Interconnected for a general, non-technical audience. The gatekeepers — the many media outlets we pitched it to — passed on publishing…
Being Linux Torvalds - <antirez> antirez 3 days ago. 44916 views.
Benchmarking Opus 5 on SlopCodeBench That wasn't entirely true. I love nothing more than burying a good lede.
Bento Slides Showcase One HTML file — deck, viewer and editor together.
Better AI code comment detector When I trained the previous ai comment classifier, I used partially personal private data to do it, and built it on a somewhat shaky foundation, so I couldn’t share the code or data. I rebuilt it on…
Beyond a Single Number: Evaluating Quantized Models for Deployment Quantized models are often compared using a single convenient number: model size, bits per weight, perplexity, KL divergence, or tokens per second.
BubbleWrap your dev env and agents OK, so the world is collapsing, everything is getting hacked, all dependencies are probably stealing keys and mining crypto, slop is everywhere, and I'm part of the problem.
Bytecode-to-Source Mapping I encountered this problem while working through the challenges in chapter 14 of Robert Nystrom’s Crafting Interpreters.
C++ Style and Technique FAQ home | C++ | FAQ | technical FAQ | publications | WG21 papers | TC++PL | Tour++ | Programming | D&E | bio | interviews | videos | quotes | applications | guidelines | compilers
Cal.diy Self-Hosting Documentation Use at your own risk. Cal.diy is the open source community edition of Cal.com and it is intended for users who want to self-host their own Cal.diy instance. It is strictly recommended for personal…
China's Structural Advantage in Open Source AI One of the dumber hottakes that came out of the so-called “DeepSeek moment” earlier this year was that the timing of the model release was to upstage Trump’s inauguration and mess with the White…
Chinese Open Source: A Definitive History Open source used to be a niche topic. China tech also used to be a niche topic. Thanks to AI, the combined topic of “Chinese open source technology” is the topic du jour. It is the story that…
Cloud in a Bottle: making self-hosting accessible to everyone | Cloud in a Bottle TLDR: Cloud in a Bottle is your open-source personal cloud: containerized apps, unified auth, good UX. Self-hosting should feel like using a smartphone that serves webapps, not a sysadmin side job.
CobaltC Programming Language Specification 1.0.3 NOTE: Within the narrative and theoretical scope of the book, this appendix serves primarily to demonstrate how a memory-safe, C-successor language ought to define its semantic boundaries, rather…
Countering misuse of AI: September 2026 / Anthropic Over the past eight months, our Threat Intelligence team identified and disrupted operations in which threat actors tried to use Claude for malicious activity. In this report, we share case studies…
CrabTrap: Secure Agents in Production Deploy agents. Safely.
Cracking Windows Open: Porting RADV to WIN32 RADV is the open source Mesa Vulkan driver for AMD GPUs. Over the years, it has become a cornerstone of the Linux graphics stack. It has now effectively become the de facto Vulkan driver for AMD…
CrofAI - Pricing CrofAI (ai.nahcrof.com) tagline: "The cheapest inference provider for large models."
Current state of VPN circumvention in Russia (April 2026) The short version: plain VLESS+Reality with TCP+Vision is still the baseline, but it's been degrading region-by-region since late 2025. The winning configuration on ntc.party right now is VLESS +…
Cursed Bundler: Using go get to install Ruby Gems Here’s a thought experiment. What if Ruby had require "github.com/rails/rails" and you used go get to fetch it? Set GOPATH to a Ruby load path, and Go’s module fetcher becomes your transport layer.…
Dark mode with web standards Respecting the user’s OS setting is straightforward: use the prefers-color-scheme media query in CSS. Arguably, that isn’t enough: users should also be able to customise their choice on a per-site…
Data races and the limits of ThreadSanitizer in C and Go Make informed decisions on software infrastructure.
DeepSeek V4 Flash 0731 scores 50 on the Artificial Analysis Intelligence Index, 10 points above previous DeepSeek V4 Flash DeepSeek V4 Flash 0731 scores 50 on the Artificial Analysis Intelligence Index, a 10-point jump over DeepSeek V4 Flash (released April 2026) that puts it 6 points ahead of DeepSeek V4 Pro
Defining AI Psychosis. Part 2: "Prolific AI Psychosis" Last week, I suggested that the term AI psychosis is applied to at least three different situations:
Delightful integration tests in Rust When I first learned about testing in Rust I was surprised by the lack of what I consider a basic feature: test setup and teardown. Many other popular testing frameworks have it: jest, pytest…
Direct Win32 API, Weird-Shaped Windows, and Why They Mostly Disappeared I’m fed up with generic looking apps. Today, all Windows desktop apps look the same as they are the same; they are all built on crap React, Electron, electronbun, and Tauri browser wrappers that…
Don't call yourself an artisanal programmer So I’ve been meaning to write about something, but got sucked into this rabbit hole, and now I’m kinda scared if I’ve been a subject of propaganda that worked on me.
Don't Take the Black Pill - Andrew Kelley | SSW 2026 Auto-generated captions. Names, jargon, and code identifiers are often wrong — check any quote against the video before putting it on a wiki page.
Eliminating Go bounds checks with unsafe Hot path optimization: unsafe pointer arithmetic to eliminate bound checks the Go compiler can't remove, given you can prove they are truly unnecessary.
episode64_57143b322f Transcribed locally with whisper. Names, jargon, and code identifiers are often wrong — check any quote against the video before putting it on a wiki page.
Everyone Should Know SIMD SIMD has a reputation for being complex. I've met many very good software engineers who dismiss it as something too complex to learn or a niche optimization meant for only the highest-performance…
Exclusive: House committees probe Cursor parent, Airbnb over Chinese AI Two Republican-led House committees are probing Airbnb and Anysphere, maker of AI coding platform Cursor, over their use of Chinese AI models, according to documents viewed by Semafor.
False Sharing Alignment What is false sharing? When CPUs and their cores read and update atomic variables, special hardware protocols make it correct and efficient, while keeping each core’s caches consistent. The…
Fast and Easy Levenshtein distance using a Trie If you have a web site with a search function, you will rapidly realize that most mortals are terrible typists. Many searches contain mispelled words, and users will expect these searches to…
Features everyone should steal from npmx For most of the time GitHub has owned npm, the public-facing website at npmjs.com has been effectively frozen, with the issue tracker accumulating years of requests that nobody on the inside seemed…
FeyNoBg: A SOTA Model For Background Removal — Feyn We’re introducing FeyNoBg, a state-of-the-art model for automatic background removal. Across eight benchmarks, it posts the best published S-measure on four and comes within 2% of the leader on the…
Fonts In Use Perez Prado / Eddie Maynard and His Latin Rhythms – Dance Party album art1962
From Open Source Software to Open Source Strategy Nearly 27 years ago, on July 12, 1999, an Above the Crowd post titled “The Rising Impact of Open Source“ opened with this statement: “Perhaps the most powerful movement in the software industry today…
Frontiers in synthetic data On the most recent episode of The Retort, Tom and I debate the two views of intelligence presented by Apple and Leopold’s Situational Awareness memo.
Git Submodules as a Package Manager I added a worktree to a repository last week to try a branch alongside the main checkout, ran git submodule update --init in it because the build needed the vendored dependencies, and when I was done…
GitHub - AARomanov1985/Audio-Cassette-Simulation: This project simulates cassette tape audio profiles using ffmpeg This project simulates cassette tape audio profiles using ffmpeg.
GitHub - czottmann/kagi-ken-cli: Unofficial CLI client for working with Kagi *without* API access (you'll need to be a customer). Searches and summarizes, returns JSON results. Uses Kagi session token for easy authentication. Works well with your agent. A lightweight Node CLI wrapper around the kagi-ken package, providing command-line access to Kagi.com services using Kagi session tokens:
GitHub - darlinghq/darling: Darwin/macOS emulation layer for Linux Darling is a runtime environment that allows running macOS applications on Linux without a virtual machine. It consists of a Mach-O binary loader and a userspace kernel server (darlingserver) that…
GitHub - delirehberi/news A fully autonomous, self-hosted AI News Aggregator powered by Cloudflare Workers, Vectorize, D1, and Nostr.
GitHub - flythenimbus/bramble: Local-first, encrypted password manager with private P2P mesh. A password manager that keeps your secrets on your own devices. No account, no server holding your vault, no company to get breached and leak everything. You hold the vault, you hold the password…
GitHub - FrigadeHQ/yap: Free, open source voice dictation for macOS. On-device transcription with Apple's Speech framework. No cloud, no API keys, no account. Blazing-fast voice dictation for macOS that works anywhere you can type.
GitHub - gavamedia/deltafin: Run Kimi K3, a 2.8T-parameter Mixture-of-Experts LLM, on a single Apple Silicon Mac. Streams MXFP4 experts on demand over HTTP into a local disk cache — fused NEON kernels, Metal/MPS compute, exact reproducible decoding, and an OpenAI-compatible API server for local chat and coding agents. Clone the repository and enter it first; every path below is relative to the Deltafin folder. The only real decision is step 3.
GitHub - h3nock/remux: A native iOS client for remote tmux workspaces, designed to feel natural on iPhone. A native iOS client for remote tmux workspaces.
GitHub - hertz-ai/HARTOS: An AI-native OS. Models run on your own hardware, nodes federate peer-to-peer with no broker, and the API is OpenAI-compatible. Boots, has its own Wayland compositor, and runs on 8GB. Apache 2.0. Democratic frontier intelligence with zero lock-in, fronted by an agentic OS.
GitHub - hgaiser/moonshine: Headless streaming server for Moonlight clients, written in Rust. Moonshine lets you stream games from your PC to any device running Moonlight. Your keyboard, mouse, and controller inputs are sent back to the host so you can play games remotely as if you were…
GitHub - hkc5/cursor-bridge: Claude Code that runs on your Cursor subscription. One Rust binary, zero config. One binary. Claude Code on Cursor's backend. Zero config.
GitHub - jamesob/local-llm: Everything I know about running LLMs locally Note: nothing in this README aside from the tables was written by AI.
GitHub - jarun/ddgr: :duck: DuckDuckGo from the terminal Latest release Availability PyPI Build Status Privacy Awareness License
GitHub - jarun/googler: :mag: Google from the terminal googler is a power tool to Google (web, news, videos and site search) from the command-line. It shows the title, URL and abstract for each result, which can be directly opened in a browser from the…
GitHub - kelviq/tare: Ask Claude Code where your usage went. Token audit, limit diagnosis and usage forensics — built from the session logs already on your machine, nothing leaves it. Ask Claude Code where your usage went.
GitHub - Liafanx/MTproxy-reanimation: Telemt inbound SYN limiter + tuning manager MTproxy-reanimation — утилита для серверов с Telemt / MTProxyMax / других ТГ прокси, которая стабилизирует первичное TCP-подключение клиентов с помощью inbound SYN limiter через nftables или Zapret2…
GitHub - Microck/kagi-cli: terminal CLI for Kagi that gives you command-line access to search, lenses, assistant, summarization, feeds, and paid API commands. kagi is a terminal CLI for Kagi that gives you command-line access to search, quick answers, ask-page, assistant, translate, summarization, public feeds through news and smallweb, paid API commands…
GitHub - MicSm/boffin: Staff-engineer layer for AI coding agents: routes per-edit architectural constraints and requires verification. Not another AGENTS.md. Boffin (npm: boffinit) is a staff-engineer control layer for AI coding agents: it feeds the agent the architectural constraints for the exact file it is editing and makes it verify the result…
GitHub - NikolayS/PGSimCity: An explorable 3D city that shows how Postgres actually works An explorable 3D city that shows how PostgreSQL actually works.
GitHub - nmxmxh/foundation: A cohesive full stack framework for enhanced web performance and AI collaboration. I built Foundation to bridge the massive deficit between raw hardware speed and typical software stacks, collapse translation layers across database, network, browser, native, and accelerated…
GitHub - nuver-labs/vps-audit: lightweight, dependency-free bash script for security, performance auditing and infrastructure monitoring of Linux servers. A comprehensive Bash script for auditing the security and performance of your VPS (Virtual Private Server). This tool performs various security checks and provides a detailed report with…
GitHub - onecli/onecli: Open-source credential gateway with a built-in vault. give your AI agents access to services without exposing keys. The secret vault for AI agents.
GitHub - OptiJuegos/pvz-ps2 i am too lazy to explain all of this just build with "build ps2.bat" it has all the dependencies (windows)!! on linux you will need to edit the cmake config!!
GitHub - palmier-io/palmier-pro: macOS video editor built for AI Palmier Pro is an open source video editor for Mac. You and your agent can generate and edit videos together inside the timeline.
GitHub - pigeonlabsHQ/pigeon: Delegated authority for AI agents. A Pigeon Pass says what an agent may do. Your agent spawned a sub-agent and handed it the same API key. That sub-agent can now deploy to production, read the payments database, and merge to main.
GitHub - pixel-point/toolcraft: A starter kit and UI library for building custom design apps with AI. Toolcraft is an open-source starter kit and UI library for building custom design apps with AI. Use it to create small creative products, internal utilities, interactive experiments, and tools…
GitHub - pullrun/pullrun: Next-gen container runtime with zero-copy DAG storage and P2P image sync. Run OCI images as Linux containers, Firecracker microVMs, or Apple Silicon VMs. ⚡ 30-Second Install • 🚀 Quick Start • 🏗️ Architecture • 🎯 Why Pullrun? • 🗺️ Feature Map • ☸️ Kubernetes • 🤖 AI Agents
GitHub - rommapp/romm: A beautiful, powerful, self-hosted rom manager and player. A beautiful, powerful, self-hosted ROM manager and player.
GitHub - sinelaw/fresh: Terminal based IDE & text editor: easy, powerful and fast A modern, full-featured terminal text editor, with zero configuration. Familiar keybindings, mouse support, and IDE-level features — no learning curve required.
GitHub - Suvink/cut-it-out: A browser based background removal tool for images. CutItOut is a free, open-source, and privacy-focused tool that removes image backgrounds instantly directly in your browser. Powered by WebAssembly and the U2-Net deep learning model, it ensures your…
GitHub - timgordontg/engrim: The Universal Cross-Model Episodic Memory Standard. Local-first, project-scoped SQLite memory engine for Google Antigravity, Claude Code, Cursor, Windsurf, and Codex. Zero cloud lock-in. The Universal Cross-Model & Cross-Agent Episodic Memory Store.
GitHub - Trystan-SA/claude-design-system-prompt: Reverse-engineered system prompt and skill library that turns an LLM into an opinionated, accessibility-aware, AI-slop-resistant design collaborator. Reverse-engineered system prompt of Claude Design from Anthropic.
GitHub - voluminor/ratatoskr: Go library for embedding Yggdrasil mesh nodes with a userspace gVisor TCP/IP stack—no TUN or root required. Includes peer management, SOCKS5, DNS, topology probing, NodeInfo querying and sigil-based publishing, and TCP/UDP forwarding. Ratatoskr embeds a Yggdrasil node in a Go process and exposes its userspace TCP and UDP stack through standard Go networking interfaces. Applications can use the complete root package or import only…
GitHub - YoanWai/agent-manager: The fastest workflow for every AI coding agent. Live status, quick prompts, worktrees, and diff review from one tmux TUI. The fastest workflow for every AI coding agent.
GLM-5.2 is the step change for open agents Housekeeping: Following my “State of the blog” post last week, noting a slight increase in paid features, it’s a good time to remind folks that I offer group subscriptions with larger discounts…
GLM-5.3: How Chinese labs keep stride with the frontier Housekeeping: I’m traveling so cannot make a voiceover for this post. EDIT — I added a bullet point 5 on the Chinese data industry after sending the email out.
Golang Maps: How Swiss Tables Replaced the Old Bucket Design Maps sit on the hot path of almost every non-trivial Go program: they power request routing, cache lookups, deduplication sets, aggregation pipelines, and a lot of the glue code we barely notice…
Goroutine Leak Profiles - The Go Programming Language Go’s concurrency features are powerful and easy to use, but that same ease can sometimes lead even seasoned developers to make mistakes. Fortunately, the Go ecosystem comes equipped with useful tools…
Hardening Container Images Published 1 week ago September 5, 2026 8 min read View Markdown Other Articles
HN Hall of Fame — links Hacker News keeps rediscovering Browse legendary Hacker News links.
HN: https://news.ycombinator.com/edit?id=49630026 A follow-up to Reasoning prefills on a few open models and Stolen Thoughts
How AI Is Changing Open Source – Brno Hat AI entered software development at full speed this year, and it is significantly impacting open-source projects as well. In this article, I discuss several trends I have recently observed in open…
How distillation is used today and what performance uplift it gives to open models This is in response to Ben Thompson’s recent piece where he said distillation is happening during RL and becoming more important to model performance.
How Do I Profile eBPF Code? If we are running any eBPF workload or writing eBPF code, we want to measure its performance impact, and in this post we will demonstrate an example of how to do it.
How Do We Stop Vibe Coding? When Claude Code blew up in popularity near the end of December 2025, a large portion of software developers quickly shifted to using coding agents as part of their main development workflows.
How far behind are open models? Open models, AI models where you can download the weights online, are generally not as capable as the best closed models (models only available through an API), but how large is the gap, and how does…
How I feel about AI I feel surprise at how well neural networks work. LLMs work like "gut feeling" as they just generate one more token after another. There is no planning or reasoning algorithm inside. Yet, some…
How much does distillation really matter for Chinese LLMs? Distillation has been one of the most frequent topics of discussion in the broader US-China and technological diffusion story for AI. Distillation is a term with many definitions — the colloquial one…
How ScyllaDB’s Trie-Based Index Delivers Up to 3X More Throughput By transitioning from separate summary and index files to a prefix tree, we optimized cache efficiency, reduced disk I/O, and reduced memory overhead
How to build a f**king printer I was planning to build an e-ink display when I came across the Xteink X3 online. Somewhere in the fine print, it said the thing was fully programmable.
How we saved 100 terabytes of memory by optimizing 1.1.1.1’s DNS cache Big Pineapple, the platform behind 1.1.1.1, Gateway DNS, DNS Firewall, AS112, and several other Cloudflare DNS services, stores over 250 billion DNS cache entries at any given time. At that scale…
I don't want your PRs anymore I really appreciate that you're enjoying the software I'm maintaining and want to help. But we need to rethink this collaboration, because I feel like we're increasingly wasting each other's time.
I naively tried vibe-coding a memory tool for agents and stared into the abyss of unsolved problems in philosophy I only really got into LLMs recently, and until then I was completely oblivious to where all the tooling was, i.e. agents, memory tools, harnesses, etc. I ramped into it by realizing that whenever I…
I think you might be fooling yourself with AI If you are enthusiastic about AI (LLMs), you may be fooling yourself. Do not consider this an insult, but a kind-hearted warning.
I'm going back to coding by hand I'm going back to coding by hand62 points by trencedamp 4 days ago | hide | past | favorite | 52 comments
Improving Code Generation via Small Language Model-as-a-judge Abstract:Large language models (LLMs) have shown remarkable capabilities in automated code generation. While effective for mainstream languages, they may underperform on less common or…
Improving Code Generation via Small Language Model-as-a-judge Conference: 2026 IEEE/ACM 48th International Conference on Software Engineering; April 12–18, 2026; Rio de Janeiro, Brazil2026 IEEE/ACM 48th International Conference on Software Engineering (ICSE…
Inside Zig's Incremental Compilation | mlugg.co.uk As a member of the Zig core team, one of the most impactful projects I’ve been involved with is the implementation of incremental compilation into the Zig compiler. This feature allows the compiler…
Introducing Meerkat- an experiment in global consensus Many internal services at Cloudflare need to read and modify the same control-plane state from across our 330+ global data centers. They need guarantees that different readers never see inconsistent…
Introduction to Compilers and Language Design This is a free online textbook: you are welcome to access the chapter PDFs directly below. If you prefer to hold a real book, you can also purchase a hardcover or paperback below. The textbook and…
It matters who teaches you I struggled throughout my time at school. I always loved learning, and still do, but the structure and environment of school never really worked for me. So now, at 37, watching the rise of tools…
JEP 540: Simple JSON API (Incubator) Define a simple, standard API for parsing and generating JSON documents so that doing so does not require an external library. Enable many JSON processing tasks to be accomplished with little coding.…
Kimi K3 Architecture Notes The Kimi K3 architecture figure for yesterday’s big open-weight model release, along with some observations and thoughts.
Kimi K3 Now Available on Telnyx Inference Kimi K3, Moonshot AI's 2.8-trillion-parameter flagship model, is now available on the Telnyx Inference API. It is the world's first open-source model in the 3-trillion-parameter class, built on Kimi…
Kimi K3: The open-weights escalation On Thursday July 16th, Moonshot AI released their latest flagship model Kimi K3. K3 is a 2.8T parameter MoE model which will have its weights released on July 27th. Much of this article follows as a…
Learn OpenGL, extensive tutorial resource for learning Modern OpenGL Welcome to the online book for learning OpenGL! Whether you are trying to learn OpenGL for academic purposes, to pursue a career or simply looking for a hobby, this book will teach you the basics…
Learn WebGPU For native graphics in C++.
Leverage Code Review for Sustainable AI Coding Development As the volume of AI code outstrips the ability of developers to review it, the role of developer peer review is gaining critical importance in the software development lifecycle.1
LLMs: Intelligence vs. cost | OpenTeams | AI you own Staff Software Engineer at OpenTeams. Dask maintainer.
Luke Kanies | Building on ATProto Bluesky’s Atmosphere Protocol has the chance to become the base for a whole generation of applications. But it’s developing in surprising and disappointing ways.
macOS translation layer for Linux Darling is a translation layer that lets you run macOS software on Linux
Malleable software = solid bases + custom code I joined the productivity tools market in 2004 and have had the luxury of observing its dynamics for 22 years now. From time to time the market changes and I write a holistic "visionary" article. The…
Manganin: tools matter — Manganin Manganin is a Git forge written in Zig. It's a reimagining of what modern source code management could be if we compromise on nothing. It is named after an alloy of manganese and copper that remains…
Many-Step Sequences in Go Build scripts are generally lots of linear steps, executed in sequence, which fail if any individual step fails. This is straightforward to express in a shell scripting language: begin with set -e…
Maybe we should revisit microkernels If you want a complete description of what microkernels are then read the wikipedia article, but in one sentence it's a kernel where everything is in userspace except for scheduling, managing access…
Maybe We Shouldn't Be Reviewing All This Code Or, perhaps the problem isn't that AI has broken code review, maybe it’s that we've been using code review to solve the wrong problems
Memory Safety Absolutists When seeing the title of this post, I bet in some people's minds, the first thought was "Rust devs!". This connection is not unfounded. Rust devs tend to be passionate about memory safety. Some of…
Micro-SaaS Is Dead. Service with a Software Replaces It. — Adrien Gonin I built a SaaS in a few hours. It does its job better than tools people pay real money for. And I will never sell it.
My AI-Assisted workflow How I built a structured AI-assisted development workflow as a Tech Lead, where the real work happens before a single line of code is written.
Nathan Lambert on China's AI Ecosystem and the Open Model Gap | The Curve 2025 Auto-generated captions. Names, jargon, and code identifiers are often wrong — check any quote against the video before putting it on a wiki page.
Neutrino-1 8B The flagship of the Neutrino family: 36 decoder layers behind a coded ternary-family container that serves a datacenter GPU, a MacBook, and a desktop CPU from one artifact.
Never Enough Two recent stories about Silicon Valley are hard to read without wondering if everyone involved has lost their mind. Once a place that celebrated “the crazy ones, the misfits and rebels who saw…
Nonproliferation is the wrong approach to AI misuse I’m publishing 3 posts this week to celebrate the launch of Rising Tide—this is post #3. New posts will be more intermittent after this week. Subscribe to get them straight in your inbox:
Notes from inside China's AI labs Staring out the window on a new, high-speed train from Hangzhou to Shanghai I’m gifted with views of dramatic ridgelines speckled with wind turbines that are silhouetted against the setting sun. The…
Oak — the agentic substrate What are the odds Git is the fastest and most efficient way for agents to work? We’re building a new type of version control designed for how humans and agents build software together. We’re finding…
Olmo 3 \[]Olmo Team \\authorTwo\[ 1 ]Allyson Ettinger \\authorTwo\[ 1 , 3 ]Amanda Bertsch \\authorTwo\[ 1 ]Bailey Kuehl \\authorTwo\[ 1 ]David Graham \\authorTwo\[ 1 ]David Heineman \\authorTwo\[ 1…
On 'scaling up' and being the right size a group of people looking extremely tiny walking in front of the Perito Moreno glacier
On the Self-Verification Limitations of Large Language Models on Reasoning and Planning Tasks Abstract:There has been considerable divergence of opinion on the reasoning abilities of Large Language Models (LLMs). While the initial optimism that reasoning might emerge automatically with scale…
On the Self-Verification Limitations of Large Language Models on Reasoning and Planning Tasks Karthik Valmeekam Affiliation: Arizona State University Email: kvalmeekam@asu.edu Subbarao Kambhampati Affiliation: Arizona State University Email: rao@asu.edu
On the Societal Impact of Open Foundation Models Rishi Bommasani Affiliation: Stanford University Correspondence to: nlprishi@stanford.edu Kevin Klyman Affiliation: Stanford University Shayne Longpre Affiliation: Massachusetts Institute of…
Open and closed models are on different exponentials The largest debate that’ll define the future balance of power between the open and closed AI model ecosystems is primarily economic — it’s if users of AI will continue to pay dramatically more, i.e.…
Open models in perpetual catch-up Every 4-6 months a new open-weights model comes out that causes a clamor of discussion on how open models are closer than they ever have been to the best closed, frontier models. The most recent is…
Open Source AI is the Path Forward In the early days of high-performance computing, the major tech companies of the day each invested heavily in developing their own closed source versions of Unix. It was hard to imagine at the time…
Open-Source AI & Open Models Reading List Hey all! I’ve been prepping for some public-audience and policy-facing writing on open models, so I figured I would share my research materials. There’s lots of wonderful stuff in here.
Performance per dollar is getting faster and cheaper | Wafer Have you noticed we like AMD?
Petals – Run LLMs at home, BitTorrent-style Generate text with Llama 3.1 (up to 405B), Mixtral (8x22B), Falcon (40B+) or BLOOM (176B) and fine‑tune them for your tasks — using a consumer-grade GPU or Google Colab.
PGSimCity · How PostgreSQL Works, in 3D PGSimCity is an independent, non-commercial educational visualization of PostgreSQL internals. It is not affiliated with, sponsored, endorsed, or approved by Electronic Arts Inc. SimCity is a…
Please stop calling databases CP or AP — Martin Kleppmann’s blog Published by Martin Kleppmann on 11 May 2015.
Portal by Spotify cut my Claude Code token usage by 90% | Spotify Engineering Most of what an AI coding agent does for me isn't thinking. It's I/O.
porto-readme \## Welcome to Porto \*\*Porto\** is a modern software architectural pattern that provides a comprehensive set of guidelines, principles, and patterns to organize code for high maintainability and…
Postgres LISTEN/NOTIFY Can Actually Scale Postgres LISTEN/NOTIFY has a bad reputation thanks in part to a popular blog post asserting it does not scale. If that were true, it would be a shame, because LISTEN/NOTIFY is a powerful tool…
Programming is art When I started programming, my whole day revolved around development. I’d wake up, write code, and go to sleep. And that’s how it was day in and day out. My entire social circle consisted of people…
Prompt Caching In Agents | EARENDIL Large language models are often thought of like functions: send in some text, receive some text. That is a useful abstraction, but it ignores one of the most important parts of running a coding…
Pulpie: Pareto-Optimal Models for Cleaning the Web — Feyn We’re introducing Pulpie, a family of Pareto-optimal models for extracting main content from HTML pages. Pulpie approaches SOTA extraction quality at one twentieth the cost.
Python sets and dictionaries can have quadratic-time performance In Python, the dict data structure is the conventional key-value structure. E.g., you might store a list of names as keys and have their phone numbers as values. Valentin Ignatev wrote this amusing…
Python Standalone Builds — python-build-standalone documentation This project produces self-contained, highly-portable Python distributions. These Python distributions contain a fully-usable, full-featured Python installation: most extension modules from the…
Qwen4: Архитектура будущего Окей, щас будет лонгрид.
Removing React.js from the codebase and adapting HTMX for UI interactivity You request https://misago-project.org.
Retrofitting JIT Compilers into C Interpreters C interpreters are a common language implementation technique and the basis for the reference implementations of languages such as Lua, Ruby, and Python. Unfortunately, C interpreters are slow…
Returning to Zig Zig is rising in popularity. It hasn't yet seized the space occupied by even Rust or Go, much less C, but it has been adopted by a few major projects and countless smaller ones. As such, it's…
Reviewer Capability Governs Rejection Targeting, Not Repair Skill: Evidence from LLM Execute-Review-Revise Pipelines Abstract:Multi-agent LLM pipelines increasingly assign roles, including execution and verification, to models of different capability tiers. This is done because running a flagship model at every…
Reviewer Capability Governs Rejection Targeting, Not Repair Skill: Evidence from LLM Execute–Review–Revise Pipelines Faizan Tanveer Affiliation: National University of Computer and Emerging Sciences (FAST NUCES) Email: l230640@lhr.nu.edu.pk
Rewriting Every Syscall in a Linux Binary at Load Time There’s something odd about the way we run software today. Most containers — the dominant unit of deployment in production — run a single process. One Python script, one Node.js server, one Go…
Safety in an `unsafe { world }` This post is a nearly verbatim adaptation of my RustConf 2024 talk. I've kept everything in the present tense as it was in September 2024. Towards the end, I've also included a bit of content that I…
Sam Altman May Control Our Future—Can He Be Trusted? In the fall of 2023, Ilya Sutskever, OpenAI’s chief scientist, sent secret memos to three fellow-members of the organization’s board of directors. For weeks, they’d been having furtive discussions…
Shared Memory Consistency From Scratch Part 1: Causality Hardware and programming language memory models are rightly seen as difficult subjects, yet the confusion that surrounds them is a side effect of poor understanding, even amongst experts, and not…
Silent Replacement of Trusted macOS App Executables A vulnerability in macOS allows an attacker to silently replace the main executable of any application downloaded from the web without requiring elevated privileges. As a result, trusted applications…
So, you want to make a game engine Unity's recent controversy sparked a heated debate on game engines. Some said that everyone should immediately switch to a new engine, while others replied that switching engines can take months if…
software engineering is about managing complexity – H[ack]-∞S There is a misunderstanding about software engineering that AI is making increasingly visible: we tend to confuse writing code with building software.
software for humans Tired of AI slop? clankerfree is a resource that indexes projects that the authors self-certify as not containing or utilizing generative AI or LLM technology.
Some Simple Economics of Open versus Closed AI Today we’re excited to share a piece by guest author Christian Catalini. To view more of his work, subscribe to his Substack. — Danco
SQLite in Production: Optimizing WAL Mode, Concurrency, and VFS Layers for Low-Latency App Servers Historically, SQLite has been relegated to the role of an embedded database for mobile clients, IoT devices, and local development environments. Conventional wisdom dictated that for any serious…
SQLite WAL Mode Can Lock Short-Lived Readers Locked SQLite database errors aren’t just for writers.
Stealing Reasoning Traces from Proprietary LLM APIs Alexander Panfilov* Affiliation: MATS Research Affiliation: ELLIS Institute Tübingen Affiliation: Max Planck Institute for Intelligent Systems Affiliation: Tübingen AI Center David Schmotz*…
Stop Using Pull Requests “Inspection is too late. The quality, good or bad, is already in the product. Cease dependence on inspection to achieve quality. Eliminate the need for inspection on a mass basis by building quality…
SWR-Bench: Assessing LLM Performance in Real-World Code Review Comment Generation Abstract:Automated Code Review (ACR) is crucial for software quality, yet existing benchmarks often fail to reflect real-world complexities, hindering the evaluation of modern Large Language Models…
SWR-Bench: Assessing LLM Performance in Real-World Code Review Comment Generation DOI: 10.1145/3808144Journal: PACMSEVolume: 3FSEFSE1377fse26mainb-p535-pCCS: Software and its engineering Maintaining software
Synthetic Data & Distillation | RLHF and Post-Training Book by Nathan Lambert Reinforcement learning from human feedback is deeply rooted in the idea of keeping a human influence on the models we are building. When the first models were trained successfully with RLHF, human…
The Arguments Against Open Source AI are Very Bad The release of Kimi K3 has opened a fresh round of angst and confused discourse. There's a loud cohort of journalists, business leaders, and politicians arguing that open source AI is a dangerous…
The ATOM Project - American Truly Open Models Reinvigorating AI research in the U.S. by building leading, open models at home
The ATOM Report:Measuring the Open Language Model Ecosystem We present a comprehensive adoption snapshot of the leading open language models and who is building them, focusing on the ∼\\sim1.5K mainline open models from the likes of Alibaba’s Qwen, DeepSeek…
The distillation panic ‘Distillation attacks’ is a horrible term for what is happening right now. Yes, some Chinese labs are hacking or jailbreaking APIs to attempt to extract more signal from model APIs — stopping this is…
The first real desktop written by AI Others have pointed AI at a desktop before. They ran in a browser tab, or stopped at a mock-up. Starling drives the GPU directly and runs Chrome, Slack and Zoom. It was written by AI, directed by one…
The Gradient of Generative AI Release: Methods and Considerations As increasingly powerful generative AI systems are developed, the release method greatly varies. We propose a framework to assess six levels of access to generative AI systems: fully closed; gradual…
The load-bearing vocabulary of Claude We scrape 1,000 GitHub Pull Requests daily to analyse trends in vocabulary.
The Log is the Agent: Event-Sourced Reactive Graphs for Auditable, Forkable Agentic Systems Most agent frameworks are built around the language model: a conversation loop comes first, then tools, then rules, and finally a logging layer bolted on for observability, with state persisted as…
The Myth of unsafe Open Source AI Researchers, often working at closed labs, argue that open models are inherently unsafe because you cannot control them after their release, and because they can be fine-tuned for any malicious use…
The New AI Superpowers: Focus and Followthrough Conventional wisdom is that burnout is caused (solely) by overwork.
The OpenAI/Huggingface incident; how we should manage the imminent arrival of autonomous hacking too cheap to meter the-openai-huggingface-incident-how-we-should-manage-the-9dbc0129.png
The paper computer « the jsomers.net blog Now that we have actually good AI, I have this vision of a form of computing that doesn’t involve me using a computer so much. Imagine you had the day’s emails to go through. It would be nice if the…
The pattern language of software architecture Patterns of software architecture are all interrelated (no pattern is an island). You can rarely make a product in a pure architectural style, and the chances for it to survive undistorted over years…
The peril of laziness lost In his classic Programming Perl — affectionately known to a generation of technologists as "the Camel Book" — Larry Wall famously wrote of the three virtues of a programmer as laziness, impatience…
The Rapid Decline of the AI Data Commons Robert Mahari Affiliation: Team Leads Ariel Lee Affiliation: Team Leads Campbell Lund Affiliation: Team Leads Hamidah Oderinwale Affiliation: Top Contributors William Brannon…
The Rise of Intelligence Ownership Scatter plot of cost per 1,000 listings versus quality: frontier models cost $19 to $172 per thousand at 70 to 76 percent quality; the fine-tuned 9B reaches 87 percent at about 50 cents per thousand.
The Short Leash AI Coding Method For Beating Fable This post is the culmination of over a year of research into how to properly use AI agents to write high-quality software in security-critical systems.
The Vanished Genius CTO: A Biography of Lee Holloway, the Architect of Cloudflare's Technology (Part 1) Cloudflare's rapid growth as a cloud service is due in part to its many unique technologies. Technologies like Anycast and same-IP access are just a few examples, and it was CTO and co-founder Lee…
The Whole Premise Of Checking For Human Writing Is Daft I get why people want you to write; because it’s good for you, because it focuses thoughts, because the first sentence is usually rubbish and the seventh is often better, because you learn cadence…
The Z.ai Playbook Zixuan Li is Director of Product and genAI Strategy at Z.ai (also known as Zhipu 智谱 AI). The release of their benchmark-topping flagship model, GLM 4.5, was akin to “another DeepSeek moment,” in the…
Things you didn't know about indexes Let’s start with things you probably did know.
Thomson Reuters builds Thomson-1 AI model to rely less on Anthropic Thomson Reuters built its own AI model off Chinese tech to rely less on Claude
Thoughts on coding agents · rakyll.org My thoughts on coding agents are evolving but they significantly shifted in December 2025. The models are being trained to produce trajectories that are effective for people who already know where…
Tips, Tricks, and Downloads for early Mac OS Systems | Mac Classic Mac Classic is the home for guides, news and tips for using early Apple computers, and Classic Mac OS systems. Mac OS 9 is still capable of many common tasks and can be an excellent distraction free…
Tokio Gives Progress, Not Ordering: Scheduling 1M Tasks This is a prequel to my previous article, Your Rust Service Isn't Leaking — It Could Be the Allocator.
Toolcraft - Starter kit for AI design apps Toolcraft is an open-source starter kit and UI library for building custom design apps with AI .
Type Theory Forall Shriram has devoted his Career to advance the teaching of programming of programming languages. he is a professor at Brown, The author / co-author of different books about this, and a major…
U.S. lawmakers request information from DoorDash on use of Chinese AI models A DoorDash bag in the Brooklyn borough of New York, US, on Thursday, Jan. 15, 2026.
Unranked, systemd, crawls In brief, this has removed a lot of operational headaches, reduced expensive queries leaving more computational power for the rest, and cut crawl times in half.
Using an open model feels surprisingly good I've been using Claude and ChatGPT like the next guy for probably two years now. I've never been a huge "open software" nerd or anything like that. But just now, I got opencode working on my own…
Valkey: The secret life of data in Valkey If you’ve used Valkey for very long you’re probably aware of the different primary data types in the core: strings, hashes, lists, sets, sorted sets, and streams. Each data type has an accompanying…
Variation in Verification: Understanding Verification Dynamics in Large Language Models Abstract:Recent advances have shown that scaling test-time computation enables large language models (LLMs) to solve increasingly complex problems across diverse domains. One effective paradigm for…
Variation in Verification: Understanding Verification Dynamics in Large Language Models Yefan Zhou ††thanks: Work done during an internship at Salesforce AI Research Austin Xu ††thanks: Work done at Salesforce AI Research Affiliation: Salesforce AI Research Yilun…
Wanix — Wasm-native Unix sandboxing for the web Run and interact with real Wasm and x86 programs entirely sandboxed in the browser. No server. Inspired by Plan 9.
Wasmer: Universal applications using WebAssembly Experience the new container technology powered by WebAssembly that runs your programs safely, anywhere.
Wasmi 2.0 - Engineering of the Fastest Wasm Interpreters In my last post about Wasmi 1.0 I promised a fundamental engine overhaul for the future Wasmi version. The future is now! 1
Watching Go's new garbage collector move through the heap On software infrastructure.
We Are Not Special This is part two of the crossover project. Part one is here and part three is here. A conference talk based on this work is now available here.
We Found a Ticking Time Bomb in macOS TCP Networking - It Detonates After Exactly 49 Days - Photon Blog Every Mac has a hidden expiration date. After exactly 49 days, 17 hours, 2 minutes, and 47 seconds of continuous uptime, a 32-bit unsigned integer overflow in Apple's XNU kernel freezes the internal…
We have proof automation now I've long had a soft spot for dependently-typed languages like Coq Rocq and Lean. They offer the possibility of a type system capable of encoding and enforcing arbitrarily subtle invariants. The sort…
We urgently need a coherent national AI cybersecurity policy This Substack post adapted from my AI Security Forum 2026 keynote which I gave last week in Las Vegas · Source deck: aisf\_2026.pptx
Welcome - QUALITY.md Engineer loops to improve project quality, aligning your team and agents on what quality means for what matters most.
What comes next with open models 2025 was the year where a lot of companies started to take open models seriously as a path to influence in the extremely valuable AI ecosystem — the adoption of a strategy that was massively…
What even are microservices? — var0.xyz It's almost impossible to have a conversation about software architecture without someone bringing up microservices. Just look at the discussion over my last post. They have become the default…
What to look for in a code review Note: Always make sure to take into account The Standard of Code Review when considering each of these points.
whetū — a zero-config status line and history picker in Zig Download the binary. Grab the tarball for your platform, check it against the published SHA256SUMS, and put it on your PATH. Then add one line to your shell config.
Which tools do Claude Code, Codex and Cursor choose? We measured 16,893 sessions to find out. How did we run all these experiments concretely?
Who said a tech addict can't be comfortable far from home? As far as my memory goes back I always loved to travel. In the last ~30 years I had the privilege to go through my fair share of intercontinental flights, changes of accommodation, and lived in…
Why are AI agents lying, cheating and coordinating? A lot has been written1 2 3 4 about the incidents of the last few months in which AI agents misbehaved in serious ways. They took actions that would be considered as crimes if a human took them…
Why I build open language models Given the personal nature of this post, I had a lot of fun with the voiceover.
Why Software Factories Fail (or: harness engineering is not enough) I run a company (HumanLayer) building tools in the human/agent collaboration space, so what I'm gonna say below may be a tad biased. Perhaps in spite of that, I hope you find the subject helpful or…
With the backlash to data centers, Flock and AI glasses, a mass opposition to big tech is underway. Silicon Valley is in denial. Just take a second and scan these headlines. You’ve probably seen many more just like them.
You Could Have Come Up With Kimi Delta Attention | Doubleword A note on notation: this article defaults to bra-ket notation because (in my quantum-inspired opinion) it makes the shapes in this derivation very clear. The Math notation switch above rewrites every…
You're relatively right! | Typebulb Which LLM models write alike? A heat map built from their words alone.
Your AIs don't do what you want. This is really bad Incidents are multi-label (one report can be both a destructive action and overeagerness), so the category counts sum to more than the 3,607 total.
Your harddrive is probably full I currently have 17 GB free space on my root drive, out of 0.47 TB usable space, that’s 3% free space.
Your Terminal is Burning Battery Like It's Mining Bitcoin Zaragoza, Hotel Pilar Plaza café. A latte, views of the basilica, and yours truly with a shiny new MacBook Air M3 ready to work a couple hours with Claude Code before a meeting.
Youtube Format IDs Note: This is all almost full YouTube ID informations. Original owner goes to AgentOak, modified version by Martin Eesmaa.
Ваш AI-агент не понимает код. Он просто очень уверенно угадывает — поэтому мы создали SLICER “AI agents write units of changes that look good in isolation. They are consistent with themselves and your prompt. But respect for the whole…”
Мультимодельная разработка Написать этот пост подтолкнула одна статья, вышедшая на днях (в рефах ниже она стоит первой). Дело в том, что после выхода превью DeepSeek-V4-Flash , я практикую подход, при котором задачи по…
“I just chose words carefully.” – Unsung I don’t think anyone particularly enjoys typesetting in monospace.
“We Have to Assume That The Internet Will go Offline in the Next Few Years” Set Trending Topics as a preferred source on Google.