What we are building, testing and learning.
Every entry answers the same eight questions: what we asked, why it matters, its state, what we built, the evidence, what we learned, what it does not prove, and what changes next. Negative results are published as results.
We are testing whether a tiny decision model can handle the hundreds of semantic judgments surrounding frontier-model work — without giving it authority to act. What if the expensive AI is doing the wrong job?
Things we built to operate the company or deliver work.
Sales Desk
2026-09-11We needed a CRM that carried the customer from first conversation into fulfillment. So we built one.
RUNNING INTERNALLYThe Agent Coordination Plane
2026-09-11We got tired of moving prompts, status and context between AI systems by hand. So we built a durable control plane around GitHub.
RUNNING INTERNALLYVerified website execution
2026-09-11A commit is not completion. HTTP 200 is not completion. The change has to be seen on the live site.
BUILDINGBounded tests with a question and a measured result.
What if the expensive AI is doing the wrong job?
2026-09-17We are testing whether a tiny decision model can handle the hundreds of semantic judgments surrounding frontier-model work — without giving it authority to act.
RUNNING LABIs the cheapest model actually cheaper?
2026-09-11Lower cost per turn did not mean lower cost per completed task.
RUNNING LABConversational Operations
2026-09-11We wanted an owner to say what changed, in normal language, and have the system turn it into structured business context.
BUILDINGAudit Lab
2026-09-11We wanted the audit to prove the product before a salesperson ever asked for a meeting.
RUNNING LABStructured collection intended to produce findings.
The most interesting AI work is happening around the model
2026-09-18Seven tools and patterns we examined this week — a typed decision model, a context compactor, a narrative evaluator, a closed-loop tester, event-triggered review, hosted executors, and the connector pattern.
RUNNING LABAI Visibility Observatory
2026-09-11We would rather show no metric than relabel ordinary search data and call it AI visibility.
RUNNING LABHow owners actually manage their digital presence
2026-09-11Five fixed questions, then let the owner talk.
PLANNED RESEARCHShort, evidence-backed observations from real client work.
Start with the business, not the homepage
2026-09-11What changes when a website redesign starts by understanding the business instead of starting with the homepage?
WORK IN PROGRESSWhat browser-level QA found that green tests missed
2026-09-11A passing build is not a passing check.
RUNNING INTERNALLYTell us what is true about your business.
A few questions, in your own words — type them or talk. A person reads what you send and follows up. Nothing is published and nothing is sent on your behalf.