๐ I'm a solo dev and my card game Moonfall is finally out
๐ฒ A card game you play like a tabletop RPG
๐ง An AI Game Master decides what happens
๐ฎ Grow your deck: fuse cards to create new ones
๐ฐ Sell cards to players who donโt know your recipes
https://t.co/V73dOOzss2
@flowersslop it's easy to restart. i like audiobooks. one takes 6h on 2x speed.
i am doing this at the gym. podcasts are depressingly boring in comparison.
remember the smell of a new booster pack
the binder nobody else was allowed to touch
the shoebox under your bed
that feeling is back
play moonfall in your browser
https://t.co/Otu3pLBXbp
1/9 Current optimizers are static during training. What if the optimizer could think and react while the network trains?
We let an agent watch the training dynamics and rewrite the optimizer's source code as training goes: a reactive optimizer.
In every language and image setting we tried that has no tuned recipe, it matches or beats the best recipe one can one-shot, and it guards against loss spikes and overfitting by itself. In the near future, optimization decisions may belong to agents, not humans.
๐ https://t.co/augwW3iFsB
๐ป Try it on your task: https://t.co/NdQNekaHoc
๐งต
๐ฎ Fuse cards to discover new ones in Moonfall
๐ฐ Sell them to other players
๐ Invest profits into new discoveries
๐ Pray nobody else discovers your recipes
๐ Play it in your browser
https://t.co/Otu3pLBXbp
Most warehouse managers never see the whole floor at once. So Claude Opus 5.5 built them the view from above.
93 people, 320 robots and every truck on Black Friday, live.
moonfall is a sandbox card game, like minecraft or roblox for cards
my challenge: craft a dragon from a pebble
i did it in 4 fusions
what would you make?
play it in your browser
https://t.co/Otu3pLBXbp
Greg Brockman on No Priors
Brockman thinks there is an opportunity to end up with a much more secure world. OpenAI moved about 25% of its production engineers onto security, telling them all projects are on hold and they are just securing now. The company turned its peak intelligence model on its own infrastructure to red team it. The team found real issues, fixed them, and saturated what that model could find. The whole exercise repeats with every new model.