Decision systemsComing soon
Why automation plateaus and intelligence compounds
A structural account of what a workflow engine cannot do, and what has to be true of a system before its performance improves on its own.
Research
Lab notes.
Working notes and longer-form writing from FirstPromise Labs. Published as findings hold up, not as a content schedule.
- AI-operated businesses
- Decision systems
- Evidence-based automation
- Agent evaluation
- Durable memory
- Institutional learning
Agent evaluationComing soon
Evaluating agents by outcome, not by transcript
Notes on measuring conversational agents against downstream results, and why fluency is a poor proxy for operating value.
Durable memoryComing soon
Memory that survives the stack
On designing the record layer so that swapping a CRM, a channel or a model is an integration cost rather than an institutional loss.
Editorial standard
What gets published here.
Writing on this page is held to the same standard as the platform: a claim needs evidence, a result needs a declared measurement, and an open question stays labeled as one.
Want to be notified when we publish?
Send a note and we will let you know when the first pieces are out.