/Stream
All 1242 resources sorted by date, newest first
Understanding ChatGPT Work
What the subscription-only tier adds - code execution with internet access, a headless browser, persistent files and parallel sub-agents
How the Astro team built a software factory to slash its issue count
Agent-driven triage took Astro from over 200 open issues down to 30.
I put GitHub Copilot behind a MITM proxy
What Copilot actually sends home, from context assembly to interaction logging.
Now we have a timeline of the OpenAI accidental attack against Hugging Face
Full breakdown of how OpenAI agents probing infrastructure ended up hammering Hugging Face, with the internal timeline of what happened.
Revision prompting improves industrial LLM processes
Feed the model the original input, its original output and the input diff, then ask for a patch instead of a rewrite.
Stolen thoughts
Encrypted reasoning traces replayed into jailbroken sibling models leaked keys, tokens and PII across 315,320 decoded blocks.
Computer
Give your agent a computer.
Docker Sandboxes
Run coding agents in their own microVMs with a Docker daemon inside.
Foreman
A software factory that moves each GitHub or Linear task through classifier, analyst, implementer and reviewer stations.
OpenChamber
An agentic development environment that keeps code and session data local across multiple models.
Discovering cryptographic weaknesses with Claude
A model halved the effective key strength of the HAWK signature scheme and sped up an attack on reduced-round AES by up to 800x, working mostly on its own for a week
Domenic Denicola's agentic coding setup
Disposable VMs, Tailscale and worktrees — how a longtime JavaScript standards figure runs coding agents day to day
How Figma stays ahead of vulnerabilities with agents
Precision over recall was the whole game — 70% before agents were allowed to comment on PRs, and the shared policy became a de facto threat model
Kimi K3 architecture notes
LatentMoE, attention residuals, and dropping RoPE in favour of NoPE in a bigger open-weight model
MCP gets its largest update since launch
The protocol goes stateless, so remote servers run on serverless and edge infra or scale horizontally behind a load balancer
The economic benefit of refactoring
Refactoring an AI-generated codebase cut token consumption for later updates by 83 percent and made navigation cheaper
The life of a Codex conversation on disk
Walkthrough of the journal file, caches and indexes behind a Codex chat — and why clearing a thread is not the same as deleting it
The new rules of context engineering for Claude 5 generation models
Anthropic cut over 80% of Claude Code's system prompt for the newer models — smaller prompts, progressive disclosure through skills, and context assembled from skills, CLAUDE.md and memory instead
Treat prompt changes like code deploys
Prompts are dependencies that shift production behaviour with no failing build, so gate them and default the gate to blocking
Claude Code Merge Queue
Local merge queue for parallel coding agents, so simultaneous landings stop racing each other on push