Keep pulling the thread on ChatGPT Images 2.
OpenAI launched its ChatGPT Images 2 model earlier this week.
In a test to recreate the game F-Zero, GPT 5.5 generated a fully functional game with three AI bot competitors and a boost feature, which the speaker described as an incredible result.
The speaker believes OpenAI's primary competitive advantage is its access to compute.
The speaker claims that Anthropic is currently 'really compute constrained'.
The speaker believes the combination of the Codex application and the GPT 5.5 model is the current state-of-the-art tool for agentic coding.
The developer tool company Cursor was recently acquired by XAI.
ChatGPT's 'projects' feature now allows users to add Google Drive files and other uploads for context.
In a comparative test, Opus 4.7 provided a more specific health insight ('build leg muscle') than GPT 5.5, which gave generic advice.
In a test, the Opus 4.7 model revealed the speaker's personal health data, stating his legs are at the 8th percentile for total mass while his body is at the 65th percentile.
The speaker believes GPT 5.5 respected his privacy prompt better than Opus 4.7, which shared personal information it was instructed not to.
The speaker assesses that GPT 5.5 is catching up to Opus in front-end design capabilities, though Opus 4.7 still has a slight edge due to more subtle animations.
In a comparative test, Opus 4.7 failed to generate a functional version of the game F-Zero, producing a 'janky' result with no AI competitors.