Writing
Notes on building an agent that acts for someone
Not explainers. There is no shortage of writing about what an AI agent is, and a model can already tell you. These are the parts we had to decide ourselves and could not find written down honestly: where the approval boundary goes, what actually breaks in autonomous agent loops, and how to charge for work that software does on your behalf. Every claim about our own product maps to code you can check.
Posts
Why an agent should ask before it ships
2026-08-03 · 5 minThe approval boundary is not a missing feature — it is the load-bearing design decision in an agent that acts on your behalf. What it costs to build, and what it buys.
What breaks in autonomous agent loops
2026-08-03 · 7 minSeven failure modes that recur across autonomous agent platforms — phantom completions, credit burn on failure, unauthorised action, credential concentration — and the design response to each.
How to charge for work an agent does on your behalf
2026-08-03 · 5 minPer-seat pricing breaks when the software does the work. Where to put the charge, why the attempt is the wrong place, and the idempotency problem nobody mentions until it bills someone twice.
On cadence
There isn’t one. These get written when there is something worth writing, which is an honest quarterly rather than a missed weekly. If you want the practical material instead, the channel playbooks are the plays our channel engine works from, and the free tools need no account.