Brain organoids, veganism, feminism, sometimes research & parenting life. Ruthless cosmopolitan. Critics rave: "Least annoying bad takes I've seen all week"
Cambridge, MA
Joined December 2019
- Tweets17.5K
- Following646
- Followers2.3K
- Likes228K
Because sentient animals evolved to pursue their own flourishing, not our instrumental goals, any workflow that exploits them as biological machines can in theory be made more efficient, reproducible, and humane by replacing the animals with designed-to-purpose synthetic systems.
Seeing "we did this for them" victory laps, but more accurately they did this for us.
If Americans hadn't put their bodies on the line in January and proved they'd rather die than let masked paramilitary feds tyrannize their cities, we wouldn't have had fair elections at all
My actual Q here is whether the exceptional generativity of African-American music is driven by fusion (eg Euro dances + West African rhythm=ragtime), or some deeper originality emerging inside Black culture.
Black cowboys—a ~1/4 minority—underperforming suggests more the latter
There's at least a certain dignity in the American political system ripping itself apart over whether to keep supporting a foreign genocide. It was humiliating when discourse about this had to be laundered through some kind of opinion about college students camping on quads
Don't fall for the edgy hype: this is not "growing human brains in labs to power AI", or anything like it.
They're squinting at simple, long-established 2D neuronal culture techniques to "get inspiration" for incremental algorithmic improvements to regular open-source AI models
today, we introduce a little mischief to the world.
for four years, we’ve been brewing real brain cells on electrodes to understand how they compute. now, we’ve turned those discoveries into a video model anyone can use.
from 0.5¢ a second, mischief 1 is at least 5x faster than its base and the cheapest image-to-video model on @fal and @OpenRouter. soon to be on @awscloud marketplace.
it’s not a delusion anymore. biocompute is real, commercial and this is just the beginning.
happy halloween.
This is in fact a positive update for me precisely because it's not very impressive.
One of the scariest things rn is that LLM training seems ~100,000x worse than biological learning, such that as soon as RSI can close the gap everything goes foom: arxiv.org/pdf/2506.14863
These results don't show this won't work, but they're doing the obvious first step (which remains hard and absurdly expensive)—growing neurons, recording how they process data, and seeing if this helps AI learn better
And this is the best they can claim: tbc.co/post/biological-neura…
