aashni⚡️

applied ml @apple, data sci + astrophysics @ucberkeley, first hire @hyperspell , ai research @nasa @stanford

[aashni⚡️ @aashnijoshi_]

oh to be in nyc rn

[aashni⚡️ @aashnijoshi_]

my friend’s doing really cool stuff 🥳

[Raghav Mundara @raghavmundara05]

Most companies fail. People rarely do.

That one line is why I've spent the last 3 months building Preflop, a new financial instrument that lets you invest in a person's trajectory.

A person raises capital by issuing tokens against their own future earnings. They pay a share of their income for a set period, then convert to a small permanent stake that trades continuously. Once the income share phase ends, the payments stop and the tokens reprice as the individuals’s career unfolds.

Issuers raise without debt or dilution. And backers get in before the world catches up, converting early conviction into shared success.

We've built markets for companies, real estate, royalties, and revenue. Never for the thing that generates all of them.

The math holds (all 350 pages of it). The pricing engine, raise mechanism, and a dozen other specs are already live. After 947 LOIs and 86 meetings across 5 cities, I realized 30 minute conversations don’t do justice to the depth here.

So I'm building the rest in public, and iterating with your feedback.

Live at https://t.co/BaKHHG7a0w

[aashni⚡️ @aashnijoshi_]

@raghavmundara05 insane stuff 👑👑

[aashni⚡️ @aashnijoshi_]

@raghavmundara05 🙌🙌🙌🙌

[marisa @meshtimes_]

30 mins into the claude code keynote and every speaker so far has been a woman. just saying 🫶🏻

@asvora @angjiang @katelyn_lesse @_catwu Dianne Penn @claudeai

[aashni⚡️ @aashnijoshi_]

@Aartiutwani 🤞

[aashni⚡️ @aashnijoshi_]

@Aartiutwani still pending approval i believe :(

[aashni⚡️ @aashnijoshi_]

@fazalmittu_ astronomy 128 data science lab, 11/10 recommend

[aashni⚡️ @aashnijoshi_]

same semester. two very different classes.

using CNNs for galaxy classification in one and writing a 4 page paper on my favourite music artist in another

berkeley classes are something else

[aashni⚡️ @aashnijoshi_]

treating college like im collecting pokemon cards

✅research lab
✅yc startup
✅faang
❌drivers license

[aashni⚡️ @aashnijoshi_]

spent thursday night learning about @vercel workflows! built and deployed a durable ai agent in typescript with the engineers who made the thing :)

[aashni⚡️ @aashnijoshi_]

sd + la last month!🐒
i already miss being on vacation

[aashni⚡️ @aashnijoshi_]

@ayushjha__ it was 5 of us haha

[aashni⚡️ @aashnijoshi_]

happy saturday to all who celebrate

[aashni⚡️ @aashnijoshi_]

the original sandbox escape was me at 4 making a break for the swings. no email sent. researcher (mom) located me visually

[aashni⚡️ @aashnijoshi_]

how i will be spending that first paycheck

[aashni⚡️ @aashnijoshi_]

@garrytan imagine being the chipotle cashier watching someone pitch their startup to a random customer

[aashni⚡️ @aashnijoshi_]

@physical_int here's the full research blog! https://t.co/AvUooqomD3

[aashni⚡️ @aashnijoshi_]

robots can fold your laundry now but couldn't plug in a charger until two days ago

came across this piece by @physical_int : their foundation model can already fold laundry and make coffee. but tasks that need sub-millimeter accuracy, like aligning a tiny screw with a screwdriver, are a different problem. the model knows what to do but can't nail the exact movement

so instead of retraining the whole model they compress its internal state into one token and train a tiny separate policy to fine-tune just the precise part. big model handles the broad strokes, tiny model handles the last millimeter

the robot ends up faster than a human operating it remotely. on ethernet cable insertion half the attempts beat every human demo in the dataset. from 15 minutes of real world data

basically it learned fine motor skills faster than me learning to use chopsticks (and i've had 20 years)