Yes, yes, yes! I asked ChatGPT to summarize my website profile from https://porg.dev, and it found and repeated the secret honeypot message that I'd forgotten about months ago.

It's the little delights in life.

I turned off my phone lock screen at home without consideration of the toddler. Now I have to go through all my apps to find errant activity 🤦‍♂️.

So far I found some autocorrected gibberish on an NBA Instagram post and a surprisingly context-appropriate 'ha' on a text thread.

From the API docs: to "turn off" DALL-E 3 prompt rewriting, use the following prompt prefix:

"I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:"

Upload an image and ask ChatGPT to "Describe this image in EXCRUCIATING detail."

It really impresses on you how well it understands.

Interesting to see tech bros talk about embeddings as if they're some novel new thing. They're older than the web!

... I'm starting to empathize with how annoying I must have been to old profs when I was an eager but naive college kid.

Monaspace - a monospaced font family with a couple of neat tricks - texture healing and ligatures - to make it much more readability while staying in the grid.

https://monaspace.githubnext.com/

Eager to try this out.