weeklyfoo #151 is here: your weekly digest of all webdev news you need to know! This time you’ll find 36 valuable links in 5 categories! Enjoy!
🚀 Read it!
- Agent Memory as a Moat: How persisting and scoping agent memory turns stateless models into systems that compound context over time by Redis / ai,agents / 17 min read
📰 Good to know
- Maximizing the Value of Your Claude Code Sessions: Clear context between tasks, set model and effort early, use @-mentions so tokens spent go to the actual task by Anthropic / ai,tools / 29 min read
- Auto-Research with Codex and a 232x Faster Kernel: Long writeup on hitting a 232x speedup on a batched Householder QR kernel with blocked algorithms, profiling loops and model-driven iteration by Sankalp / ai,performance / 29 min read
- A Quick Look at Zero-Knowledge Proofs: How a prover convinces a verifier without handing over the solution, and where ZKPs help beyond crypto, like age verification by Max Bernstein / security,crypto / 12 min read
- Everything Is About to Go Dark: On AI-written software, its security fallout, and what happens to lawful access to communications by Matthew Green / security / 15 min read
- Something Nobody Told You About the Image Element: The img element is container and content at once, so object-fit and border-radius can make it overflow in surprising ways by master.dev / css,frontend / 9 min read
- Protobuf Finally Has LSP Support: Buf shipped the first full Language Server Protocol implementation for Protobuf, so proto files get real IDE support by Buf / tools,protobuf / 5 min read
- Compression Is Prediction: Compressors and language models turn out to be solving the same problem, walked through from the basics by Annie Sexton / ai,deepdive / 24 min read
- Red Agent Exploits Snowflake Vuln Missed by Copilot: An autonomous agent found and exploited a GitHub Actions flaw five days after it was introduced by Wiz / security,ai / 9 min read
- Git at Any Scale: Why Git packfiles break down on a server, and the stateless S3 write-ahead-log architecture Cursor built to get linearizable pushes by Cursor / engineering,git / 60 min read
- Consistency Is the New Latency: Agents reading stale replicas build whole reasoning chains on wrong facts, and writing those back creates hallucination debt by AWS / databases,ai / 13 min read
- I Am Worried About a Prompt Injection Worm: Semi-autonomous agents roaming the internet with too much authority are the AI security problem to watch by Daniel Miessler / security,ai / 6 min read
- CSS, the Bomb Inside Your Inbox: How CSS in email clients is ripe for exploitation, with a pile of working attack techniques by Gareth Heyes / security,css / 51 min read
- PagedAttention, Virtual Memory for the KV Cache: Borrowing OS paging for the KV cache so a shared pool of fixed-size blocks cuts waste and lifts throughput by Gustafson / ai,performance / 20 min read
- GitHub, Autoscaling, and the Component Substitution Fallacy: The outage came from an autoscaling policy that ignored the Istio sidecar limits, not from one broken component by Lorin Hochstein / devops,engineering / 6 min read
- How We Built Our Write-Ahead Log: Long writeup on building a WAL on object storage, and why 60-year-old System R ideas suddenly matter again by Chroma / databases,engineering / 24 min read
- Introducing the Half-Day: AI-found exploits arrive faster than any human researcher can produce them, and what that does to the vulnerability market by Margin Research / security,ai / 10 min read
🧰 Tools
- Code Storage: API-only Git storage from ex-GitHub engineers, already backing Lovable and Bolt for user-generated code by Pierre Computer Company / git,api
- oj: A bundler with Vite’s API and performance that uses far less memory by Rapha Amorim / bundler,javascript
- Mole: Deep-research agent that works under an expense budget and verifies its claims with citations by lajosdeme / ai,agents
- AI CSS: Free customizable UI components for AI agents, covering thinking states, tool calls, streaming text and structured output by aicss / ui,ai
- Morphicons: Morph any stroke SVG icon into any other, closed-form rotation plus spring physics, zero dependencies by Guillermo Lopez Gomez / ui,icons
- Saggar: Native macOS terminal manager that surfaces which shells, tests and coding agents are waiting on a decision by Marginal Utility / macos,cli
- TermDOM: Build terminal UIs and interactive CLIs with plain HTML, CSS and DOM APIs by TermDOM / cli,javascript
- MicroLighter: Small client-side syntax highlighter built on the CSS Custom Highlights API by Dave Rupert / css,javascript
- Origin: A git forge for the agentic era, from the Cursor team by Cursor / git,ai
- Needle: 14MB foundation model for tool calling on tiny devices like phones, wearables and robots by Cactus Compute / ai,models
- Tegaki: Generate stroke data from any font and render handwriting animations in React, Svelte, Vue, Solid, Astro or vanilla JS by gkurt / animation,javascript
- Bun 1.4: 1,500 more Node.js tests passing, five times less CPU, 35% less memory, plus Bun.Image, Bun.WebView and HTTP/3 by Bun / javascript,node
- SvelteKit 3.0 RC: Vite 8 with Rolldown, error boundaries, validated env vars and remote functions by Svelte / javascript,frameworks
- Preact 11.0 RC: Adds use(), createPortal() and useEffectEvent() among other things by Preact / javascript,frontend
- Astro 7.2: Incremental static builds, and you can now opt out of session support by Astro / javascript,frameworks
🤪 Fun
- A Daily Doodle Website as a Joyful Middle Finger to AI Art: Bakatako is a browser-based daily doodle site where art has to be drawn on its own canvas, so everything is human-made by Creative Bloq / fun,art / 16 min read
- Firefox in WebAssembly: There is nothing to see here, it is just a browser running in your browser by Puter / wasm,fun / 1 min read
📺 Videos
- Creator of TypeScript on the 10x Faster Compiler: Anders Hejlsberg on why Go over Rust, how little the rewrite leaned on LLMs, and whether AI replaces juniors by Ryan Peterman / typescript,ai
- From Chrome DevTools to AI Engineering with Addy Osmani: 14 years at Google, the risk of cognitive surrender, and why an AI model can never be accountable by Gergely Orosz / ai,career