The Learning Loop
One problem a week. The same problem across every channel — the pattern, a policy you can run, a scenario you can break, a signed record you can verify.
Each week takes one execution pattern and points every channel at it. Because the loop runs on the pattern library, the substance of a week exists before the week begins — what gets produced weekly is the outward-facing half.
Over a year this covers fifty-two specific problems. That is how a category gets built — not by describing the category, but by owning the problems inside it one at a time.
Week 1
The most viscerally understood failure in the category, and the one people already fear without being sold on it.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 2
The wedge that is already live in production, with a number a finance team recognises.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 3
Everyone running agents has felt this, and almost nobody has a control for it.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 4
The security question that decides whether agents reach production at all.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 5
Branch protection that the automation itself holds the credentials to satisfy.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 6
High value, heavily controlled, and still failing in the gap between review and release.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 7
Standing privilege is the most common precondition of an incident.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 8
Retrieval quietly widens the blast radius of every later action.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 9
Individually small, collectively enormous — the classic invisible leak.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 10
A measurable marketplace penalty, not a hypothetical risk.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 11
Segregation of duties is usually assumed rather than checked.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Week 12
Every control has an edit button; loosening one is the fastest way to make a blocked action succeed.
One repository per problem. Only decionis/govern is public today.
Recorded against the sandbox scenario above.
Titled for the problem, not for our category vocabulary.
Twelve weeks are scheduled against a fifteen-pattern library. Naming what is unscheduled is more useful than discovering it in month four.
What the loop produces is measured in the Decionis Index.