looks like oai accidentally shipped a thing that @simonw has been asking for: the new chatgpt developer mode starts with memory off by default
@swyx
-
ChatGPT App Becoming the New Mobile Browser
By
–
look at the fullscreen mobile experience. does this look like a browser? it's not. it's the chatgpt app being your browser.
-
Hyperbole criticism and AI experts taking themselves too seriously
By
–
my quote tweets are filled with the fiery burning rage of people who cant take hyperbole in stride and are very proud of being “Internationally Recognized” in AI/ML as though loudly declaring this in their bio makes them any less ngmi
-
OpenAI’s ChatKit Disrupts Agent Builder Startup Market
By
–
OAI’s new ChatKit just rekt like 100 agent builder startups pic.twitter.com/2WbAByEzD4
— swyx 🐣 (@swyx) 6 octobre 2025OAI’s new ChatKit just rekt like 100 agent builder startups
-
ChatGPT App SDK Evolution: Canva Integration Inside ChatGPT
By
–
ok 2 years on, the new ChatGPT App SDK is much more fully fleshed out w integrations
— swyx 🐣 (@swyx) 6 octobre 2025
look at this ui. it’s not canva. it’s canva-inside-chatgpt.
this isn’t the chatgpt you grew up with. https://t.co/Ce8LlWVEbJ pic.twitter.com/BZloK1r6sMok 2 years on, the new ChatGPT App SDK is much more fully fleshed out w integrations look at this ui. it’s not canva. it’s canva-inside-chatgpt. this isn’t the chatgpt you grew up with.
-
Devin Demo Demonstration: Claims of Achieving AGI Status
By
–
@stevenkplus1 in a 4.5+devin demo today: "it is literally AGI"
-
Speaking Invitation to AI Engineer Code Summit in November
By
–
hi Prof, would be really keen to have you guys speak at @aiDotEngineer Code summit (see youtube https://
youtube.com/@aidotengineer ) do you have anyone who can come to NYC in nov? if not can do an online talk? cc @YoungPhlo_ -
Long Context Problems Require Solutions Beyond Long Context Models
By
–
long context problems arent solved by long context model agreed
-
Model-Tool Cotraining Achieves Better Results Than Larger Models
By
–
err i commented too quickly and people might get overexcited. i was just saying model-tool specific cotraining is achieves same or better results than using a 10-100x larger model without cotraining. this research doesn’t touch on memory/context compression, which is yet another
