So I did my first Substack live last week with Exploring ChatGPT and honestly it went way better than I had any right to expect.
We spent damn near an hour tearing into the stuff that keeps people up at night. Prompt injection as an unsolved class of vulnerability, how data leaks happen when you let an LLM chew on your docs, the alignment problem as a security problem, and the dark art of guardrail bypass.
The interview format actually worked. Exploring ChatGPT asked the right questions, the kind that force you to stop hand-waving and explain the actual exploit chain. If you missed it, the recording's up. Go watch it, come back, tell me where I'm wrong. That's how this works.
Thank you Clint C., Farida Khalaf, Jules, Will Ty, and many others for tuning into my live video with Exploring ChatGPT! Join me for my next live video in the app.
