From what I have understood so far, ChatGPT Dots is like adding an extra orchestration layer between the human and the agent (Codex). Instead of going back and forth with Codex chat, you can tell your high-level idea to your Dot and it will do that on your behalf.
What's the difference between ChatGPT Dots and Codex? Maybe I'm missing something, but they seem kind of similar. Maybe it's OpenAI's take on OpenClaw?
I hope people realize that models are now so good that you don’t need to obsessively create skill md files for every security use case. You should do the opposite and ask the model to follow high-level principles instead of something specific, especially for security use cases. #appsec
I've seen so many people comparing AI to the industrial revolution, but I think it's a false equivalence. AI can’t be compared to the industrial revolution or anything else; this is the first time in human history a technology can match or exceed human intelligence.
It's embarrassing to admit, but I found out today that by default Claude Code deletes chats after 30 days of inactivity. That's why I sometimes couldn’t find chats in the terminal, which I thought was just a UI bug messing with the chat history.
I've been using Claude Code in the terminal for so long, but I've finally seen the light and switched to the desktop app, which is way more convenient in every way.
Yeah i take my words back, the patch that they applied is even more hilarious, i don’t think anyone from security team is even involved because how else would you explain this😂😂
The Meta AI support assistant account takeover is hilarious. How could their security team have missed this? But I also empathize with them—working at such a huge scale and getting everything absolutely right is not always easy.