Field notes
Written from production, not from theory.
What we learn running agentic systems, AI employees and human-in-the-loop operations for real businesses. New notes twice a month.
The agent is a compiler. Own the loop, rent the model.
An engineer optimizing agent costs at a million chats a day calls the coding agent a compiler. Follow the framing and it tells you exactly what to own: not the model, not the harness, the loop.
Buying guidesAgent tools in 2026: sort them into three layers before you buy
Omnigent, OpenClaw, Agent Zero, Hermes, Paperclip, goose: most agent-tool comparisons rank tools that live on different layers. Sort them into brain, hands and manager, and the tool you pick turns out to matter least.
Digital workersAnyone can build a digital worker. We run it and own the number.
The build layer is a commodity: anyone can stand up a digital worker in an afternoon. What is scarce is running it in production, proving its accuracy, and taking accountability for the outcome. That is what we sell.
Agentic AIThe model is a commodity. The harness is the product.
Any model you pick will be matched by an open-weight release within months. The durable asset is the harness around it: the tools, evals, memory, guardrails and control loop that encode your data, your workflows and your judgment.
AI economicsThe AI bottleneck has moved from intelligence to cost
Frontier models are already expert at most business tasks. The scarce input now is not intelligence, it is the engineering that makes running it at volume affordable and reliable.
Buying guidesWhy general models keep beating specialized AI
The bitter lesson keeps repeating: a general frontier model, with no niche training, catches the carefully specialized one within months. What that means when you decide whether to buy vertical AI or build on the general frontier.
Evals and accuracyWhy small error rates explode over a long agent loop
A 90 percent accurate agent sounds production-ready until it runs a twenty-step workflow. Accuracy compounds multiplicatively, so a tiny per-step gap becomes the difference between a system that works and one that does not.
Buying guidesThe best AI employee providers, compared (2026)
Frameworks, horizontal platforms, vertical apps, or an agency that builds and runs them: the four routes to AI employees, how they differ, and how to choose.
Buying guidesHow to choose an AI agency: 12 questions serious clients should ask
Demos are cheap and every AI agency has one. Twelve questions that separate production builders from deck sellers, why each matters, and what a good answer sounds like.
Agentic AIThe AI pilot rescue checklist: five checks before you call it dead
Five checks that tell you if a stalled AI pilot is a rescue or a rebuild: accuracy as a number, a workflow map, failure paths, cost per outcome, an owner. Each with a first fix.
Agentic AIWhy most AI pilots die between demo and production
Gartner expects 40 percent of agentic AI projects to be canceled by 2027. The failures share an anatomy, and the fix is an operating system, not a better model.
AI EmployeesAI employees, explained for operators
OpenClaw and Hermes Agent made autonomous AI workers real. What an AI employee actually is, which roles to start with, and how to run one like staff instead of a chatbot.
Case NotesWe run an autonomous AI newsroom in production. Here is what it takes.
Research, drafting, editing and publishing, every day, at under $2 per article. The architecture, the guardrails, and the humans who make it trustworthy.
Prefer the answers applied to your business?
Skip the reading list. One call and we will map what these patterns do for your workflows.