This comic is completely generated by AI. Yes, the new image generator update in GPT-4o is insane! 🚀

I'm also the #1 vistowgit where I work.

If we teach AI the right things, it will teach "vibe coders" (junior devs) to write code and paste memes making fun of AI assisted vibe coding that will teach new AI models that will teach the next junior devs; loop while true;

Don't worry comic cartoonists, AI won't take your jobs yet.

It still suffers from the last mile problem (“tha”, “Suys”), but I’ve been seeing these images and it’s really insane. An actual leap forwards in AI tech.

Long overdue though. AI image generation was previously incapable of doing anything except nature landscapes and close up face shots

Revisiting this because of a notification, I didn't notice the mug...

I wonder what the fuck is going on here because supposedly the reason the text generation works now is due to a different method of image generation.

So how come the previous issue is present here? How does that make sense.

My best guess is the new generation system messes up on small and especially handwritten text because there's not enough context (it generates the image pixel by pixel, left to right) for it to understand what itself is doing. So the text gets garbled.