When Software Updates Arrive as Instructions
When software updates arrive as prompts, the signature authenticates the instruction — not the program the local LLM produces from it.
Category: Development
The Kanban board is the file: lightweight coordination for parallel Claude Code
A Kanban board tells you is what's about to happen, what another agent is doing right now, or which card is yours to pick up.
Building a development diary for Claude Code
Commit messages tell you what changed, but don't tell you what was tried first, what broke, or where the work pivoted. A diary does.
Skills Are Harness Engineering You Can Do in a Markdown File
A skill is a packaged set of instructions that gets injected into a context when it is relevant. In its simplest form, it is a SKILL.md file.
The Agent Harness: Everything Except the Model
Harness.is everything in an AI agent that is not the model itself.:"If you're not the model, you're the harness."
Agents Addressing Agents
Multi-agent communication is an active design space. Google shipped A2A as a dedicated agent-to-agent protocol.
Human code review is dead
Somewhere in your engineering org right now, there's a pull request that has been open for four days.
Code as Action: The Pattern Behind Programmatic Tool Calling
There's a design decision buried inside every AI agent that most developers make without realizing they've made it: what is the atomic unit of action?
A Practical Guide to Getting Started with OpenClaw
Somewhere in early February, Andrej Karpathy — founding member of OpenAI, former AI director at Tesla, a person not easily impressed by internet phenomena…
Vercel’s Eval Data Suggests Static Context Beats Skill Retrieval for AI Coding Agents
There's a recurring theme in AI engineering that never gets old: the "dumb" approach beating the sophisticated one.