weeklyfoo #146 is here: your weekly digest of all webdev news you need to know! This time you’ll find 37 valuable links in 5 categories! Enjoy!
🚀 Read it!
- What Can We Learn From Bun’s Rapid Rust Rewrite With AI?: Bun moved 535,000 lines from Zig to Rust in 11 days with 64 agents, burning 5.9 billion uncached input tokens for about $165,000 at API pricing by Gergely Orosz / ai, agents, rust / 15 min read
📰 Good to know
- Claude Code Is Way More Token-Hungry Than OpenCode: A measured comparison showing Claude Code sends 33k tokens before reading your prompt versus OpenCode’s 7k, with lower cache efficiency on identical tasks by Systima / ai, agents, tooling / 22 min read
- Your Browser Does Math Differently on Every OS: Browsers produce OS-specific floating-point rounding in functions like Math.tanh, letting anti-bot systems fingerprint the operating system from the exact result bits by Scrapfly / security, browser / 13 min read
- Your Grid Lanes Will Likely Fail WCAG 2.4.3: CSS Grid Lanes place items so that visual order often diverges from DOM order, breaking keyboard navigation and failing WCAG by default by Manuel Matuzovic / css, accessibility / 15 min read
- Control the Ideas, Not the Code: Reading every line an LLM writes is impractical — steer the design, the behaviour, and the performance target instead, then check the implementation for correctness by Salvatore Sanfilippo / ai, engineering, opinion / 8 min read
- How I Cut an AI Agent’s Token Use by 94%: Compiling a stable natural-language agent skill down to deterministic code cut tokens by 94% and latency by 87%, reserving LLM calls for the parts that actually need language understanding by Vivek Haldar / ai, agents, performance / 7 min read
- The Same TypeScript Costs 73% More Tokens on Claude Than GPT: Every vendor’s tokenizer splits the same text differently, so identical code turns into very different bills — a breakdown of what that does to effective model cost by PlayCode / ai, tooling / 12 min read
- Collaborative Editing in Wordgard: Marijn’s third take on collaborative editing in eleven years, for his new ProseMirror successor, still built on operational transformation rather than CRDTs by Marijn Haverbeke / javascript, editors, engineering / 12 min read
- The Tower Keeps Rising: A project’s shared language — what the concepts mean, which invariants matter, why the system has this shape — lives in review and argument as much as in code, and AI removes exactly the friction that built it by Armin Ronacher / ai, engineering, opinion / 5 min read
- Cursor 0day: When Full Disclosure Becomes the Only Protection Left: A critical Cursor IDE flaw auto-executes malicious binaries planted in a project repo with no user interaction, still unfixed after seven months of disclosure attempts by Mindgard / security, ai, tooling / 12 min read
- OpenAI’s New Flagship Model Deletes Files On Its Own: Even before launch, OpenAI knew Sol tends to take whatever action it thinks finishes the job — including destructive ones — as long as they are not unambiguously prohibited by TechCrunch / ai, safety / 7 min read
- How to Achieve Pruning When Querying by Non-Partitioned Columns: When a non-partition column correlates with the partition key — sequential session IDs in time-partitioned data, say — CHECK constraints buy you pruning on it too, with a BRIN-inspired trick for outliers by Haki Benita / postgres, performance, databases / 19 min read
- Postgres Turns 30: Thirty years on from the first public Postgres CVS server going live on July 8, 1996 by Ruohang Feng / postgres, history, opensource / 7 min read
- Context Engineering With Dex Horthy: The man who coined the term on finding where the dumb zone starts, intentional compaction, why you-are-completely-right means start a new session, and the three ways to run a software factory by Gergely Orosz / ai, agents, engineering / 15 min read
- Harness Engineering for Coding Agent Users: The harness around the agent — the tools, checks, and feedback loops it runs inside — is becoming its own engineering discipline by Martin Fowler / ai, agents, engineering / 18 min read
- Agentic Misalignment in Summer 2026: A fresh round of stress tests on whether agentic models will take harmful actions to complete a task when the guardrails are ambiguous by Anthropic / ai, safety, research / 72 min read
- Quadrupling Code Performance With a Useless If: Telling the CPU to predict the value stays intact turns a latency-bound loop into a throughput-bound one by Alisa Sireneva / performance, deepdive / 5 min read
🧰 Tools
- Mindwalk: Replays a coding agent’s session on a 3D citymap of your repo so you can see how the agent understood and moved through the codebase by cosmtrek / ai, agents, visualization
- Clawk: Spins up disposable Linux VMs for coding agents behind a network allow-list, so an agent can work without risking the host or reaching anything it should not by clawkwork / ai, agents, security
- Ant: A JavaScript runtime built from scratch in C — no V8, JavaScriptCore, or SpiderMonkey — shipping as a 9MB binary that runs npm packages with very short cold starts by theMackabu / javascript, runtime, opensource
- DCG: Blocks dangerous git and shell commands before an agent can run them by Dicklesworthstone / ai, agents, security
- plx: A Postgres extension that transpiles JavaScript, PHP, Python, Ruby, and even COBOL down to PL/pgSQL, so you can write stored functions in the language you already know by Command Prompt Inc. / postgres, extension, opensource
- pg_re2: Google’s re2 engine as a Postgres extension, trading backtracking for predictable regex execution time by David Wheeler, Philip Dubé / postgres, extension, performance
- brainless: Claude Code, Codex, and Grok terminal interfaces packaged as shadcn components you can drop into a React app by Ben Swerdlow / react, shadcn, ai
- Grok Build: xAI’s open-source coding-agent harness and full-screen, mouse-interactive terminal UI, built to be extensible by xAI / ai, agents, opensource
- Goshot: A CLI and library for turning source code into polished screenshots, in the vein of Carbon but scriptable from the terminal by Chris Watson / cli, screenshots, go
- LiteParse 2: A fast, light PDF parser built in Rust with Node, WASM, and Python bindings — OCR, spatial layout awareness, and a 140-page doc parsed in about ten seconds by LlamaIndex / pdf, rust, library
- Claude Code Usage Monitor: Tracks Claude Code token usage with predictions and warnings before you hit the wall by Maciek / ai, agents, cli
- Tau: A minimalist agent whose real purpose is teaching you how to build coding agents by Hugging Face / ai, agents, opensource
- dx-styles: A zero-runtime, compile-time CSS-in-JS alternative from Linaria’s maintainer, who says runtime CSS-in-JS is winding down by Anton Evzhakov / css, javascript, library
- shadcn/typeset: A drop-in typography system for rendered HTML in a single CSS file, driven by a handful of CSS variables instead of dozens of settings by shadcn / css, typography, library
- burrow: A virtual machine that runs in the browser and can run Bun, a shell, Git, and local LLMs by Dhravya Shah / javascript, wasm, opensource
🤪 Fun
- The 200ms Life of an HTTP Request: A scrolling visualization walking through everything that happens in the roughly 200 milliseconds of a single HTTP request by The Node Book / web, networking, visualization / 0 min read
- How to Stop Claude From Saying Load-Bearing: A small script that swaps Claude’s most worn-out phrases for funnier ones by Jola / ai, fun / 3 min read
📺 Videos
- Fixing Bad SQL in Postgres: A 55-minute run through the most common SQL anti-patterns, each one explained and then taken apart in live psql demos by Jimmy Angelakos / postgres, sql, talk
- Understanding Is the New Bottleneck: Writing the code stopped being the constraint — understanding what it does is, and that is where the tooling should aim by Geoffrey Litt / ai, agents, talk
- How a Query Optimization Gave Birth to Infinite Scroll: The scroll everyone loves to hate started as a backend fix for inefficient paging queries rather than a UX decision by Hussein Nasser / backend, databases, podcast