Security·10 min read·
The Command Said It Worked
Five commands this site published reported success while doing nothing. A small container lab reproduces four, and the cheap defense is running each command once where it must fail.
Vol. MMXXVINo. 98
Field reports from cloud-security engineering, AI experiments, and a homelab that keeps getting bigger.
Latest writing from William Zujkowski
Start here: agentic security, tools, and homelab reading paths
Security·5 min read·
One ZIP, Two Lists of Files
A small reproducible ZIP lab shows why indexed and streaming readers can inspect different entries in the same archive.
Linux·5 min read·
Two Processes, One Page Cache: Testing Shared State Before Blaming the VM
A small Linux experiment separates shared file identity from identical bytes, with a failed tmpfs control and a careful look at a July 2026 VM-isolation preprint.
Security·7 min read·
A Correct Answer Doesn't Mean the Agent's Memory Is Clean
A synthetic memory-recovery drill shows why a correct task result and a clean derived state are separate checks.
Security·5 min read·
Your Backup Administrator Should Not Be Able to Delete Yesterday
Timelock Drive puts retention below a compromised host. An offline model shows why unfreezing a backup must start a countdown, not erase its protection.
Security·6 min read·
The Database Gets a Vote on Your Read-Only Agent
A disposable PostgreSQL lab tests what an agent-facing database role can actually do, including the SELECT that writes through a privileged function.