Exists now
- A personal loop that captures correction candidates.
- Human review before a lesson reaches operative instructions.
- A durable correction history used under real working pressure.
Standing
Standing captures correction candidates and puts review between repeated feedback and the instructions an agent follows. The agent can surface a lesson. A person decides whether it deserves to become a rule.
The problem
The same correction keeps returning because the lesson never earns a home.
A useful correction often disappears with the session. The next agent repeats the mistake, receives the same feedback, and loses it again. Automatically promoting every correction is worse: the agent would gain permission to rewrite the rules that govern it.
Standing keeps the candidate, the review, and the eventual instruction as separate events.
Intended user
People who maintain persistent agent instructions across real projects.
The useful signal is recurrence: a correction appears again, survives review, and becomes a rule that future sessions can actually find.
Current boundary
The mechanism is active. The distributable application is still taking shape.
The larger application remains a north star; completing every hardened V1 concern is not a website launch gate.
Still open
- The smallest useful external release.
- A supported installation and storage contract.
- Portability across agent hosts and instruction targets.
Availability
In development. The personal loop is not a public product.
Standing will change status when another person can use a bounded release without inheriting the author’s machine and assumptions. There is no public build, price, or waitlist today.
Public receipts
No distributable release or public trial yet.
The current loop proves the mechanism for its author. The next receipt needs to prove a useful boundary for someone else.
Writing
The work behind Standing
Design decisions and future release stories will be published on Sea Trials when they have public evidence attached.