OPENAI ROADMAP UPDATE FOR GPT-4.5 and GPT-5: We want to do a better job of sharing our intended roadmap, and a much better job simplifying our product offerings. We want AI to “just work” for you; we realize how complicated our model and product offerings have gotten. We hate the model picker as much as you do and want to return to magic unified intelligence. We will next ship GPT-4.5, the model we called Orion internally, as our last non-chain-of-thought model. After that, a top goal for us is to unify o-series models and GPT-series models by creating systems that can use all our tools, know when to think for a long time or not, and generally be useful for a very wide range of tasks. In both ChatGPT and our API, we will release GPT-5 as a system that integrates a lot of our technology, including o3. We will no longer ship o3 as a standalone model. The free tier of ChatGPT will get unlimited chat access to GPT-5 at the standard intelligence setting (!!), subject to abuse thresholds. Plus subscribers will be able to run GPT-5 at a higher level of intelligence, and Pro subscribers will be able to run GPT-5 at an even higher level of intelligence. These models will incorporate voice, canvas, search, deep research, and more.

That is, during training, we used Answer Token Prediction on 80% of samples, and applied normal MLM (random 30% masking) on the remaining 20% of samples. Except I coded it wrong, and it actually assigned a single meaningless label to all masked tokens. The loss for those...

just furnished my living room guys how do you like it

McDonald's employee with down syndrome retires after 32 years of serving smiles.

Today I completed a quarterly, full-day customer support rotation, a program that every exec & VP at @vercel is enrolled in. If you got a reply from a nice fella named 'Martin' today, it was yours truly 😁 One fascinating learning is that our strides in support AI & @v0 have had a non-obvious effect: many of the cases were hard. They required a lot of analysis, came from large production codebases, and involved lots of realtime data. For us, this actually means we'll double down on Support & Field Engineering. As AI takes care of the easy stuff, support cases that make it through will contain more and more high-quality signal that inform product changes and framework design improvements (I ended up writing down a couple dozen ideas over the course of the day). Soon you'll get our CTO and CPO, so… give us the best you've got 🫡