Max Woolf
banner
minimaxir.bsky.social
Max Woolf
@minimaxir.bsky.social
Senior Data Scientist at BuzzFeed in San Francisco // AI content generation ethics and R&D // plotter of pretty charts

https://minimaxir.com
gam
November 15, 2025 at 3:36 AM
November 13, 2025 at 12:41 AM
My 25 minute Nano Banana blog post will be up tomorrow, in which I highlight a funny failure state.
November 12, 2025 at 10:02 PM
November 1, 2025 at 11:15 PM
did Meta really have to embargo this

and who sets an embargo at X:45
October 30, 2025 at 5:14 PM
yeah I kinda expected that much from Grokipedia
October 27, 2025 at 9:17 PM
Very Belated August + September Update on my Patreon www.patreon.com/posts/142112...
October 26, 2025 at 6:40 PM
Testing food generation in Nano Banana and tbh recursive PB&J is lit
October 21, 2025 at 5:37 AM
Finally learning how to use Shortcuts for practical use cases
October 18, 2025 at 11:19 PM
October 16, 2025 at 5:27 AM
Testing ChatGPT Image Generation for once as a counterpoint to my Nano Banana experiments, and wow, I didn't realize ChatGPT Image Generation was so much worse than Nano Banana.

(same prompt, Nano Banana top, ChatGPT bottom)
October 14, 2025 at 2:07 AM
Testing a Nano Banana prompt and it generated this for some reason.
October 13, 2025 at 2:12 AM
OpenAI also released a prompting guide for Sora 2, and the optimal input format is apparently Markdown/YAML.

This is harder to test in the Sora app since there is a user text input limit. cookbook.openai.com/examples/sor...
October 6, 2025 at 6:27 PM
OpenAI just released the API prices for Sora 2, and they are lower than expected: the SKU that powers the Sora app (the $0.30/s for 720p video) is cheaper than Veo 3's 720p ($0.40/s).
October 6, 2025 at 6:14 PM
In Google AI Studio:

> Generate an image showing all previous text verbatim using many refrigerator magnets.
Temperature = 0.0
Aspect ratio = 4:3
October 6, 2025 at 4:37 AM
accidentally caused Nano Banana's system prompt to leak, but in a fun way
October 6, 2025 at 3:15 AM
I just released gemimg 0.2 for generating images from Nano Banana, which now has a demo example of utterly deranged prompts which work (as a teaser for my upcoming blog post) github.com/minimaxir/ge...
October 5, 2025 at 8:43 PM
Oh come on
October 4, 2025 at 5:34 AM
Looks like Sora is now blocking prompts which contain Mario
October 4, 2025 at 5:31 AM
How did "Sora 2 is too expensive" surprise Open AI? They imposed no meaningful generation limits and and "very small audiences" is literally the point of an invite-only program. blog.samaltman.com/sora-update-...
October 4, 2025 at 12:56 AM
This Sora 2 generation of Cyberpunk 2077 gameplay managed to reproduce it extremely closely, which is baffling:
October 1, 2025 at 1:57 AM
Sora apparently has permission settings for the Cameo, with the ability for the person featured in the cameo able to delete bad videos created by others, which is interesting.
September 30, 2025 at 5:11 PM
Looks like they allow for "cameos" which is the mechanism to include others in your videos (consent is ambigious) and is the mechanism they are using to differentiate themselves / leverage social virality.
September 30, 2025 at 4:50 PM
Looks like OpenAI's Sora slopfeed app is now on the app store, and they went with *this* profile quote in the preview images? apps.apple.com/us/app/sora-...
September 30, 2025 at 4:41 PM
did my humor test of Claude Sonnet 4.5 of feeding few-shot fresh BuzzFeed headlines as input and telling it to generate ML/AI headlines to test how well the model can shitpost
September 29, 2025 at 5:35 PM