Training by voice note
You don’t tune the system with settings. You tune it the way you’d correct a sharp new hire — by telling it what was off.
You reply to the morning brief with a voice note: “Too long.” “That’s off-voice.” “Never lead with price.” “Use my words for this, not that.” Every correction lands back in the-static-brain as a rule it keeps. The next batch comes back closer to how you’d have written it.
Why this is the whole training interface
No prompts to engineer, no dashboard to configure. Talk the way you already think about your own work, and the system absorbs it. By month two you’ve stopped re-explaining your business — it already knows, because you’ve been teaching it a rule at a time.
Why it compounds
A voice note isn’t a one-time fix. “Never lead with price” doesn’t correct one email — it becomes a standing rule every future output reads. That’s the difference between a tool you re-brief forever and infrastructure that learns you once. It’s the human half of the-compounding-loop: the agents don’t get smarter, the brain they read from does — and you’re the one sharpening it.
Related: the-static-brain · the-compounding-loop · the-morning-brief