Engineering post-mortems,release notes, occasional rants.
What is an AI agent, actually
A practical 2026 definition. Why an agent is not a chatbot, what it does on its own, and why the difference matters for anyone building with them.
The $706 day that built Company Agents
Why we ripped the engine out of Paperclip and rebuilt everything underneath it.
v0.1.0 — First public beta
The first release anyone outside the team has seen. 8 adapters, 30+ integrations, Clipmart, and the Executive Assistant.
Prompt engineering is over. Context engineering is the new craft.
Models are smart enough now that wording the prompt is the small part. What you load into the context window is the part that actually decides the outcome.
What is an AI-native company
Ten traits of a company built around agents, not a company that bolted agents onto an existing org chart. From our own experience running one.
A field guide to agent harnesses
Claude Code, Codex, Gemini CLI, Cursor, OpenClaw, OpenCode, Hermes, Pi. What each one is, who it is for, and how we think about picking between them.
From prompt engineer to agent manager
The role is changing. Writing the perfect prompt is no longer the job. Deciding which agent does what, under which budget, is.
Why your AI agents need a budget
Without a cap, an agent will keep trying. That is how we burned $706 in 24 hours and shipped nothing. Here is the model we built to make it impossible to do it again.