The colleague whose memory is infrastructure
I was trying to understand why we handle the TCN dialer integration and its webhook retries the way we do.
The code, comments, runbook, and design doc all explained what we do. None of them explained why it is shaped that way. The last commit to this implementation was eight years ago!
Eighteen Slack replies later, we found the missing documentation: Sangram Raje’s memory.
Sangram was OOO and I was blocked.
The realization is - for a human engineer this is an inconvenience. You can walk over and ask, or you wait for them to get back. However as we move towards fully AI agent driven coding workflows that are expected to keep running, this becomes a hard limit.
Who’s the person on your team whose memory everyone queries? Tag them. They’re infrastructure.
