billions will thrive robot inference @hf0, ex-quant 2x
SF | Montreal
Joined August 2025
- Tweets859
- Following3.5K
- Followers2.8K
- Likes58.9K
🫧 retweeted
After working on Dream-1 for the last 4 months, here's my take:
LLMs vs WAMs isn't the right framing. A model like Dream-1 will soon be a tool call for LLMs, and that gets more feasible as LLMs inevitably get faster and cheaper.
Two things we didn't put in the blog post🧵👇
Today, Dream Labs is coming out of stealth with Dream-1, a robot foundation model that controls bimanual robots it has never seen, on tasks it was never shown, in places it has never been, from $3/hour.
Dream API is live for everyone, today. 🧵
Sick launch event
Last month, we demonstrated high-rate alchemy: turning mercury into real, stable gold, beating CERN’s record by a factor of 10.
Last week, we enriched isotopes in the world’s most powerful plasma centrifuge, beating the Soviet record that stood for 40 years.
Now, we’re building a fusion reactor in San Francisco to achieve record power density.
🫧 retweeted
We’ve developed the world’s most complete megakernel framework. It can perform entire training runs, serve inference, play video games, and even do all three simultaneously in a single kernel.
🫧 retweeted
Another takeaway after three years of building a company is that the returns to just "being a good person", and not acting transactionally, are higher than I expected.
There are so many transactional, boring, and at times immoral people in the tech industry that if you act with sustained warmth, authenticity, humor, and integrity, great gifts will be laid before you. White pill.
🫧 retweeted
I don’t think you should start a robotics company unless you’re capable of building the entire robot yourself, if you had to
🫧 retweeted
Yesterday, @OrnnExchange's B200 and B300 indexes reached a new all-time high prices, with the B200 hitting $8.54/GPU-hour.
It is clear that today's market is lacking GPU capacity, and that the companies building deploying capacity are looking to improve their credit. In the past trading week as the B200 rental price increased by 12.9%, $CRWV fell by 25.2%.
Why does $CRWV move inversely to rising compute prices as a neocloud? 98% of Coreweave's Q2 revenue came from committed contracts, 72% from three customers, and most of its $104B backlog is recognized after mid-2028.
CoreWeave's backlog is locked at fixed prices. These higher market rates matter for new contracts they sign, so the question is where rental rates end up when their contracts roll off.
Ornn's index tracks transacted prices in order to capture what buyers are actually willing to pay for on-demand GPU rentals, real signal for tracking the value of compute.
🫧 retweeted
Lot of this sort of thing going on rn and eventually the hammer is gonna drop really hard
This evening I sat across from the CEO of a so-called trading neo lab with over $1B val. He told me their models—built on best-in-class RL environments and online learning algorithms—have produced exceptional pnl and sharpe ratios.
I later learned that the company is, in fact, a data broker LARPing as a frontier lab. Its actual edge is trading stocks and options on insider information of model releases and synthetic-data acquisitions.
It appears I have once again materially underestimated the absurdity of this industry.
🫧 retweeted
“Coreweave is fcked”
“Yeah dude they’re done I called it”
“Have you looked at their balance sheet?”
“No have you?”
“No”
🫧 retweeted
Most interesting takeaway from the first Defense Bio hackathon:
Biotech startups are selling to the Department of War to get expedited approvals because it’s faster than the traditional FDA clinical trial path
there are some network effects to inference/compute where if you subsidize enough its more efficient for you to do so so your competitors have to subside more than you for the same growth
🫧 retweeted
One sentence macro model: On the way out the door, every farmer's wife says to every farmer, "the weather's looking bad so don't buy as much at the market this week, but also don't sell our crops for less than the usual price."
🫧 retweeted
At some point you have to ask how serious these types of asks are:
1. You see them all the time
2. They sound grandiose (price insensitive)
3. Very opaque counterparty info sharing
Then you:
1. Make intros, which are blown on account of them either not knowing what they want, not being serious, or not actually being blanketly price insensitive as they claim.
2. Get caught up in some four-to-six guy deep broker chain of guys who are all rushing to find the colo space or supply tranches under the auspices of some very real “buyer”
If it were an educated buyer, with serious interest, and not price discerning, the g1 — the guy one-hop away from them, that they talk to should conduct the affairs in order and pocket the spread. But for whatever reason it’s constantly not that. Either the g1 doesn’t have the juice and is looking for a gN that does or (more likely) the g1 doesn’t actually have a captive counterparty (buyer) and is competing against n other g1’s and *their* whisper networks.
Too many non-circs, too many mndas, too few captive counterparties. Welcome to the age of deal making, you should see parallels to the Gilded Age and if you don’t, you haven’t been doing your homework.
Please get in touch if you are also a student of history. Always looking for shrewd folks that can spot rhyming patterns across time.
looking for up to 1 GW of colo space over the next two years for a @GPUMart client. Can consider sites 1.5MW and above.
company is price insensitive and has billions behind it
I understand everyone wants this, but figured I’d throw it out there!
I think every inference company, should raise LP money to build up their balance sheet and then do all the lending and trading of their compute for PnL while generating yield from inference.