/Stream
All 1219 resources sorted by date, newest first
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
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
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
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
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
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
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
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
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
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
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
brainless
Claude Code, Codex, and Grok terminal interfaces packaged as shadcn components you can drop into a React app
Claude Code Usage Monitor
Tracks Claude Code token usage with predictions and warnings before you hit the wall
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
DCG
Blocks dangerous git and shell commands before an agent can run them
Grok Build
xAI's open-source coding-agent harness and full-screen, mouse-interactive terminal UI, built to be extensible
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
Tau
A minimalist agent whose real purpose is teaching you how to build coding agents
An Off Switch for Dual-Use Knowledge in AI Models
Anthropic's GRAM (Gradient-Routed Auxiliary Modules) isolates dual-use knowledge like cyber offense/defense into dedicated model compartments, resisting attempts to recover removed knowledge
GitLost: How We Tricked GitHub's AI Agent Into Leaking Private Repos
An indirect prompt injection through public GitHub issues let attackers exploit GitHub's Agentic Workflows to extract data from private repositories