Keep pulling the thread on Mark Chen.
Meta has approximately $10 billion of capital per year to deploy towards talent acquisition.
OpenAI employees are willing to stay at the company for compensation packages that are multiples below what Meta is offering.
OpenAI allocates more compute to exploratory research aimed at finding the next paradigm than to training the final model artifacts.
OpenAI made a bet more than two years ago to focus on applying reinforcement learning (RL) to language models, a decision that was unpopular at the time.
OpenAI has internal models that perform at the level of Google's Gemini 3.
OpenAI is confident it will soon release models that perform at the level of Gemini 3 and can release successor models that are even better.
OpenAI has been focusing on supercharging its pre-training efforts for the last six months.
GPT-5 Pro represents an inflection point in frontier research, capable of understanding a physicist's latest paper after 30 minutes of processing.
An OpenAI model achieved second place at the Coder competition, jumping from 100th place the previous year.
A petition to bring back Sam Altman as CEO was signed by over 90% of the OpenAI research organization by the morning after it was conceived.
OpenAI does not believe that "scaling is dead" and sees significant remaining potential in pre-training.
OpenAI is creating an initiative called "OpenAI for Science" to accelerate scientific research by providing tools to scientists.