Notes on hiring, security, and how we build Kit.
Check AI agent sandbox security beyond the microVM: verify shared files, network access, credentials, external tools, and how your team accepts results.
Check whether the Alpine OpenSSH password bypass applies to your servers, distinguish PAM and container cases, and verify the fix reached the running daemon.
Build a solo founder incident response plan before your first security report: name contacts, record evidence, rehearse decisions, and check recovery access.
Build a fuzz testing crash report another engineer can reproduce, with the exact build, input, command, failure evidence, owner, and fix verification.
Use a feature-combination matrix to test shared previews, document privilege crossings, triage vulnerability reports, and verify the complete fix.
An AI security analysis refusal leaves the review unfinished. Preserve the original artifact, record the limit, and assign authorized human follow-up.
Revoked certificates can still be trusted by developer tools. Use the eMudhra incident to build a remediation inventory and verify each affected client.
Check CVE-2026-98432 Linux kernel fixes and CVE-2026-106378 Chrome updates, including affected systems, exploit prerequisites and patch verification.
Review AI tenant isolation before sharing hiring data: separate provider promises, minimize candidate context, and test the boundaries your team controls.
Displaying items 1-9 of 33 in total