Using AI to uncover security bugs
Using AI to uncover security bugs is a hot topic right now.
Claude Code & Checkpoints
Claude Code released an update, and one of the new features is checkpoints.
Codex prompting guide from OpenAI
This Codex prompting guide has some great tips, and echos some things I’ve shared here before, in particular: split large tasks and Leverage Codex for debugging
Source Control & Vibecoding
Source control is such a useful companion for vibecoding. Code commits offer a deterministic checkpoint for progress. The determinism is important; I don’t trust AI to do the right thing when I ask it to undo a change.
Threadweaver: My Latest Pet Project
My latest pet project: Threadweaver
Being good isn’t enough
Simple and spot-on career advice from https://joshs.bearblog.dev/being-good-isnt-enough/
Leadership lessons from developing new QBs
I found this Athletic article on NFL quarterbacks fascinating, not only because I’m a Bears fan, but also because it embodies some good lessons in organizational leadership.
Vibedebugging
I had an interesting use case where vibecoding just wasn’t fixing a pesky bug. So I tried “vibedebugging” instead, and used the root cause to build the correct solution.
My experience on 9/11
These 13 photos are all I have left from 9/11, out of maybe 200+ I took that day. These photos remain because I posted them and some thoughts to my Xanga site. The other 200+ photos lived on a hard drive that crashed a few years later. A useful reminder to always have backups.
Prompt injection and early 2000s blogging
“A weakness in OpenAI’s Connectors allowed sensitive information to be extracted from a Google Drive account using an indirect prompt injection attack.” (source)