sujayalluri retweeted
77% of ChatGPT messages are prompts a laptop can answer locally. 95% of its users pay nothing. The industry's answer is $100B+ data centers.
The AI world is splitting in two, and the next great AI company gets built in the gap.
sujayalluri retweeted
The greatest trick OpenAI ever pulled was convincing the world that sacrificing your privacy is the price of using AI.
We think that's insane.
Introducing Loci: your own private AI, inside the apps you already use.
→ text it from iMessage, WhatsApp, Telegram, Discord and Slack
→ runs 100% on your own computer
→ chats never leave your machine
Try it free: askloci.ai/chat
sujayalluri retweeted
I was previously in the camp of "SWE will be solved". Instead, the field will bifurcate.
Great SWEs are more demanded than ever: AI amplifies their value.
Those with the ability to craft solutions LLMs are incapable of regardless of model/harness: low-level optimizations, clever system design, maintain increasingly complex codebase, good product taste, and experiment design sense are in a great spot.
SWEs who just executed on plans that others laid out are at risk. Those who have historically been in the latter group can definitely cross over, though.
sujayalluri retweeted
Wow. Less than a month ago, Jev released and decision models took over X. 3 weeks later, @liquidai shipped a 3B that goes head to head with Jev -- on a Mac, 30 ms latency, and *with* vision...
The frontier is reaching local hardware faster than anyone could've anticipated.
Today we release Open d1: two open-weight multimodal models in our d1 decision model family.
> d1-3B: text + vision
> d1-omni-600M: text + image or text + audio
> Real-time decision making anywhere, from data centers such as @nvidia DGX to RTX workstations to Jetson at the edge.
1/
sujayalluri retweeted
Many in tech blowing this off because it didn't push fwd the frontier while neglecting the implications it has on cheapening near-frontier intelligence. You don't always need the newest toy.
Awesome to see American open weight models catching up.
Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.
- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch
Full weights release this month.
Learn more about Beam: reflection.ai/beam
sujayalluri retweeted
Local AI caught up way faster than anyone realizes. Sharing my thoughts having spent time in ML at Meta and dedicating the last year to getting smart on local AI.
1/ Barely anyone, not even hardcore AI power users, realizes how fast local models have caught up to the frontier. Qwen3.8 woke a few people up, but the most technical people are so frontier-pilled they've blinded themselves to the breakneck progress.
Meanwhile, normal users don't care about frontier intelligence at all. They want AI that gets the job done, but they have no idea what a local model is or how to run one.
So local is stuck in a weird spot: the people who can run these models don't care to do so, and the people who'd benefit most don't know how to run local.
sujayalluri retweeted
can confirm knowing at least 3 ex-quants (T1 shops) making high 6 / low 7 figs trading prediction markets
paying very low taxes on gains (form llc to trade)
living at parents' penthouses (cuz its Manhattan no one has a basement)
among some of the smartest, most independently-minded, and interesting individuals I know
and also got plenty of time to hit the gym, spend time with family & friends, and go on dates as well
NY residents are collectively up more than $200 million in net profits on @Kalshi
the way I see it, traditional sports books charged egregious spreads for people to gamble
prediction markets allow competition to compress that spread, meaning people can bet for cheaper
and the profits from liquidity provision goes to ANYONE who's willing to warehouse risk smartly
instead of being concentrated in the hands a few corporations who ban anyone who's good
yes excessive gambling is bad
but as someone who trades perps, options, pred mkts, and memecoins for entertainment (and also perfectly knows how these markets work)
I don't think its the government's job to tell me, or any other adult, what they should do with their money
yes I know that some of my trades (especially fun options punts lmao) are negative expectancy,
but I actually don't think I can get such a powerful surge of dopamine anywhere else for cheaper
financial markets are competitive and efficient, and so risk is provided at the LOWEST possible cost
and as for the people who I am "giving" that edge to, I'd rather it be some dude my age building a trading bot than some multi-billion dollar cooperation
founders like @mansourtarek_ has allowed for ever greater opportunity for high-agency people, this is fundamentally a good thing
“They are in their parents’ basements making markets,” Fitzsimmons said. “That’s where some of the explosive growth is coming from.”
wsj.com/finance/kalshi-predi…
sujayalluri retweeted
it’s not a real founder dinner unless the boys have their laptop open to generate shareholder value during dinner
sujayalluri retweeted
Today @Brendan_McCord and I are launching Cosmos Ventures.
AI is overturning assumptions society has organized itself around for centuries. The models keep getting better, but no level of capability can tell us what kind of society we should make with them.
We need a new class of founders to build what comes next.
Instead of only “building what people want” and iterating their way to product-market fit, these founders start from the question of how we should live together. They build new institutions that set the standards for how entire industries operate. They combine a philosophical vision of the society they’re building toward with the technical ability to realize it.
We call them philosopher-builders.
Immanuel Kant gave the Enlightenment its motto: Sapere aude. Dare to know.
The AI age demands another.
Condere aude. Dare to found. 🫡
sujayalluri retweeted
it's a pretty unique experience tbh. i grew up in louisiana where no one knew how to code and now i just show up and get carried by the cog team all day
I asked people which companies have the highest density of talented people they know:
1st. Cognition - 9 votes
Equal 1st. Anthropic - 9 votes
2nd. Modal - 5 votes
3rd. OpenAI - 4 votes
4th. Standard Intelligence - 3 votes
4th. Cursor - 3 votes
Two votes each:
- Ramp
- Flapping Airplanes
- DeepMind
- Long Lake
- Applied Compute
One vote each:
- SpaceXAI
- SpaceX (treated separately, one vote was for the AI lab subsidiary and one was for the rocket team)
- American Terawatt
- Mechanize
- Olix
- Fluidstack
- Chai Discovery
- Sail Research
- Etched
- Core Automation
- Specter
- Clay
- Applied Intuition
- Sierra
- Hivemind
- Bitrig
- Retro
- Thinking Machines
- Decagon
- Precigenetics
- Pangram
- Reflect
- Thrive Holdings
- Adaption
sujayalluri retweeted
I found the best new chicken sandwich in NYC. It’s from Old Mate’s Pub this Aussie bar in Fidi (I know random) but it’s a 24 hour marinated fried chicken sandwich on a seeded brioche bun, it’s actually incredible. Idk why no one’s talking about it.
Might be the most underrated food item in NYC rn. Calling it rn gonna be the new sandwich of the city.
sujayalluri retweeted
damnlines points cameras at the most line-plagued restaurants in NYC and LA and tells you how long the wait actually is before you leave the house. Right now it's watching NYC spots like L'Industrie Pizzeria, John's of Bleecker, and Salt Hank's, counting every person in line in real time so you never have to gamble on a two-hour wait again. Here's the breakdown:
> Currently tracking lines at L'Industrie Pizzeria (142 watching, ~9 min wait, 82 in line), John's of Bleecker Street, Salt Hank's, and Breakfast by Salt's Cure
> Entire system is built on Little's Law, a 60-year-old queuing theorem: L = λ × W (queue length equals arrival rate times wait time)
> Wait time for joining right now is calculated as West = L / μ, where μ is the service rate measured by watching how many people actually exit the line over a rolling window
> When there isn't enough data for a reliable service rate, it falls back to a conservative per-person estimate so you're more likely to be pleasantly surprised than stranded
> People counting pipeline runs computer vision on each video frame, draws bounding boxes around individuals, then tracks identities across frames so person #42 stays person #42 even after they move
> A queue zone is defined in the camera's field of view and the count increments on entry, decrements on exit (current_count = Σ entries − Σ exits)
> Dwell time per person feeds directly back into the service rate calculation, so the estimate gets sharper the longer the system watches a given spot
> Fresh estimates get pushed to your screen every few seconds throughout operating hours
sujayalluri retweeted
Back is January, Microsoft’s CEO Satya Nadella told me: "All I know is, I'm good for my $80 billion."
Well, he just upped the ante this afternoon.
He is now “good for” $120 billion.
sujayalluri retweeted
I don’t think Vlad Tenev gets enough credit.
Unreal glowup over the past 4.5 years.
sujayalluri retweeted
The most vocal critics of the debt and deficit tend to be people who run successful trading businesses: Dalio, Griffin, Druckenmiller, Singer, Solomon, Gross, Dimon, Gundlach, etc. Here's my theory why...
Every successful trading business is grounded in risk management. Long-term success depends on avoiding systemic blowups. These managers are confident their platforms will thrive over time, but only if they steer clear of tail-risk events and so they're acutely focused on avoiding them.
The US is the most successful platform the world has ever seen. This country has delivered enormous gains in living standards and health and has tremendous potential to continue doing so, but only if it avoids the tail-risk events.
Excessive debt is one of those. It poses an existential risk for a country. Nobody can predict when, how, or even if excessive debt will trigger a crisis, just as no trader can predict when a poorly managed position might implode.
That uncertainty is no excuse to ignore the risk. That's what bothers traders about the national debt. The US is holding a position with significant and potentially catastrophic risk for little benefit. And instead of managing it responsibly, it keeps adding exposure.
In trading, ignoring risk is the most reliable path to collapse. The same holds true in governing a country.