AI agents & reliability
Your AI agent said “Done”. Was its state actually saved?
A completed run and a recoverable state are two different promises. A controlled fault in the real Mecatl engine makes that boundary visible.
Read the noteTechnical notes
How I think
about systems.
A few questions that matter in migrations, delivery and operations. Each note follows one line of technical reasoning.
Migration & operational risk
The way back belongs before cutover.
Traffic is often easier to switch back than data state. A credible migration plan makes that distinction visible before it becomes urgent.
Read the noteGitOps & diagnosis
What a green sync does not prove.
A status is only as useful as the question it answers. An operational decision also needs to account for the questions left unanswered.
Read the noteOperational data & diagnosis
The dashboard is not the investigation.
A new view can look familiar and still lead to a different conclusion. This note collects general questions for comparing operational data.
Read the noteThese notes explain my technical perspective. Illustrative examples are identified as such. Confidential project details are not published.