Red Black Tree Blog
Srđan Marković
9 min read
A governance-first operating model for AI in engineering: exit ramps, capped spend, and a gated program that stops if the numbers do not hold.
Series, 2 parts
Running AI under governance
Two parts on adopting AI in an engineering organization without losing control of risk, cost or quality, and on measuring what it changes.
Part 1 of 2
Running AI under governance: an operating model for engineering organizations
Srđan Marković
9 min read
A governance-first operating model for AI in engineering: exit ramps, capped spend, and a gated program that stops if the numbers do not hold.
Part 2 of 2
Is "percentage of code written by AI" a good metric? What to measure instead
Srđan Marković
8 min read
Is percentage of code written by AI a good metric? It is not. Keep review gates identical, then measure DORA-class metrics and escaped defects.
Series, 3 parts
Secret management for developers
Three parts on keeping developer credentials out of reach of AI coding agents, from a macOS Keychain pattern to 1Password CLI and headless SSH sessions.
Part 1 of 3
Secret management in the age of AI coding agents
Srđan Marković
4 min read
Remove plaintext API keys from shell configs before AI coding agents read them. A macOS Keychain pattern for zero-trust dotfiles.
Part 2 of 3
Scaling secret management with 1Password CLI
Srđan Marković
4 min read
Scale the zero-trust shell config pattern across multiple machines with 1Password CLI, Touch ID, and batch injection via op inject.
Part 3 of 3
SSH and the boundaries of macOS security
Srđan Marković
4 min read
Solve the SSH secret loading problem on macOS where Keychain and Touch ID fail remotely. Using 1Password Service Accounts for headless sessions.