Approving a rule once, trusting it forever
JournalTorus
How we make automated fixes explainable
Written by Owen Strand, Product Lead
Friday, 19 September 2026
Automation only helps if people trust it. And people only trust what they can read. Every fix Torus makes starts as a rule a human approved, written in plain words, with a log entry that says what changed and why.
Rules people can read
A rule says what Torus watches, what counts as wrong, and what it is allowed to do about it. If an engineer cannot explain the rule to a colleague in one sentence, we rewrite it until they can.
A log that answers questions
Each change is logged with the signal that triggered it, the rule that allowed it and the result a minute later. When someone asks why a server restarted at four in the morning, the answer is already written down.
What every rule contains
- A trigger in plain words
- A limit on how often it may run
- A way to undo the change
- The name of the person who approved it
Key takeaways
- Trust comes from reading, not from accuracy alone
- Small, reversible fixes first
- The log is part of the product
How we make automated fixes explainable.
