Max Uper (@uper_max) header
Max Uper (@uper_max) profile picture

As above, so below

no w he re
Joined March 2022
Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)Photo by Max Uper (@uper_max)
Max Uper retweeted
"b300s are so cooked" "facts bro" "provider just offered me $4.6/hr, can take as many as I want with no commit" "sheesh" *hits vape* "deadass got 64 nodes online in Kazakhstan in 2 weeks, DM if you need more capacity" "alright bet" 👍
12
10
3
246
17,760
Replying to @marclanders
You can't spell healthcare without "THC."
9
25
351
24,253
Max Uper retweeted
We've developed a new inference engine, called 'e'. It's the closest thing to alien technology which we have, and pushes 10,500 tok/s aggregate on a single DGX Station with the new @NVIDIAAI Nemotron 3.5 Super VL model, released a few minutes ago.
128
151
87
3,053
380,062
The outcome of this constraint is many more small clusters. To turn non-IG credit into IG, you need diversification. That’s the alchemy that unlocks large swaths of capital. Not 1bn to one borrower, 1bn to 100 borrowers. Then you can forget the Nvidia backstops and ig wrappers, and turn GPUs into a standalone financable asset class. Standardization is what makes ai legible to Wall Street. The parallel constraint in the power market is already pushing us towards 2-5mw deployments instead of 250mw. Both of these forces will be positive for the sustainability of the industry, as small deployments with a couple of month lead time can better match supply to actual, current demand, reducing the risk of overbuilding. Ironically, over time we might even see these smaller deals pricing similarly to larger ig deployments, despite worse credit, as the project complexity becomes a major tradeoff if there is an abs market to sell into.
The AI buildout needs more capital than backstops can cover. “There’s bank lenders, insurance lenders, they don’t want to take a lot of risk. You’ve got private equity, they want to take a lot of risk. You’ve got private credit, they’re in the middle.” “What you do is you tranche risk. If there are losses, the equity will take the loss first, then the B piece, and then part of the A2. The loss has to be so severe to affect the senior tranche, which is how you can create different risk profiles.” “We think NVIDIA has the capacity to support up to 46 gigawatts. That’s still not enough. There’s 240 gigawatts that’s gonna come in.” “The lending market has to evolve to take risks outside of backstops. It’s really the biggest question of our time.”
1
4
19
1,182
Max Uper retweeted
$/MWh × time-to-power is the new optimization. the generation stack drives this, with developers choosing between deep markets for modular reciprocating engines and higher-efficiency large gas turbines with much longer procurement cycles. reciprocating engines are getting interesting because they’re modular, widely available, and can start monetizing power much sooner, even though large gas turbines can have better economics at scale. when compute is waiting on electrons, time matters. that changes what “efficient” means. we need power equipment with a short delivery curve.
10
22
156
9,264
Max Uper retweeted
Just like with oil, the market outside the majors is the true marginal market for compute.
34
24
16
124
29,688
Max Uper retweeted
Where does money invested into the AI buildout actually go? For every $100 flowing into the supply chain: - $50 to chips - $20 to power - $15 to networking - $15 to cooling, buildings, and land More charts in State of Markets II: a16z.news/p/state-of-markets…
Introducing our State of Markets pt 2, along with a companion podcast where we unpack the data and discuss what comes next. Tech is the everything cycle. Supply: putting the buildout in context, just passed railroads as % of GDP. The wisdom of Elon is real: the factory (or the datacenter!) is the product. Demand: diffusion is so, so early. Median AI vendor spending in the top 1% of companies is 8x that of the top 10%. Only about 30% of S&P 500 companies report a quantified AI impact, which means there’s a substantial opportunity in connecting models to a company's data and workflows. Diffusion into companies is one of the main themes of the next 5 years. We’re entering the agent work period. Only a few million users today, but applicable to billions of internet users with massive surplus created. META/GOOG monetize US users at $200+ per year today. Agent opportunity is much higher. Mega-trends the next 5 years: Consumer agents, Robotics, Autonomy, AI x bio, Personal health, Diffusion into enterprise, New era of American Dynamism. Much more in our SoM report here - a16z.news/p/state-of-markets… @a16z @sarahdingwang @aleximm @santiago__rdz
176
867
154
4,671
1,065,329
Max Uper retweeted
The most useful idea I've come across for understanding myself is a fairly woo-woo one called Internal Family Systems (IFS). I think most of us have incorrect models of ourselves. We think of ourselves as individuals, with needs and desires that are legible, hierarchical, and rational. A bundle of desires, beliefs and preferences all moving in one direction, making up one self-concept. As nice as this might be, I suspect the reality is much of what we call "I" is a mashup of differing, random, and often conflicting desires, patterns, and experiences. IFS models your self-concept as a collection of parts, plus a core "Self" that can relate to them. Some "parts" may want success and fame, others may be scared, others may hang out and judge other parts as they see fit. I've found this "parts work" surprisingly effective. For whatever reason, it's much easier to admit to myself that there are parts of me that want to be well known, to be successful, or to be well liked. And when I can admit that, it's much easier to accept that those parts exist, without the self-judgement that comes from not wanting to admit that "I" am this way or that. It's a small (but effective) way to keep your identity small. More and more, I feel like maturing is a slow process of accepting parts of you that you wish were different. Of accepting and surrendering to the reality that - for almost all of us - there are parts that we don't love or are unhappy about. I've found IFS surprisingly helpful in working through some emotional challenges in my own life. If you're interested in learning more, I'd recommend the (fairly woo-woo) book No Bad Parts as a good starting point.
18
2
1
119
8,576
Max Uper retweeted
Are you implementing Jev into your x402 client for tool selection? Reach out. Quick thoughts in thread.
23
5
3
79
8,911
Paul is the smartest person in European venture. This is a must read!
In 1852 aluminium cost twice as much as gold. Then the price fell over 99.9%. Yet it built one of the largest materials markets on earth. What it says about AI tokens:
22
113
4
1,375
399,923
Max Uper retweeted
out of all the rich people you can hate, mfers chose the *one guy solving rocket science* instead of having orgies in the Aegean youd retire after 2M in net worth but yes, the guy choosing to still work for civilization after 1 TRILLION is the problem derptardio
180
289
29
5,301
157,704
Max Uper retweeted
This is the most unique perps exchange design I've seen in a while. Particularly because of the $PAPER mechanism. Here's how it works: > @papertrade_xyz is a synthetic perps protocol on HyperEVM without fees, funding, or slippage. > If you lose a trade, your full margin goes to the LP. If you win, you get paid but the protocol takes a fee on your gain - the fee is smaller for big moves, bigger for tiny moves. That haircut is the protocol's only revenue. > The LP starts at $0. It fills purely from trader losses. If the LP can't pay a winner immediately, the payout enters an onchain FIFO debt queue and gets paid as the LP refills. > Losing traders receive PAPER tokens. The lower the LP balance, the more PAPER you mint per dollar lost. > PAPER can be staked to earn USDC dividends - a continuous cut of LP revenue. Once the LP exceeds $5M, all surplus gains go entirely to stakers. > When the LP is low, emissions are high, so this is theoretically the time to accumulate PAPER cheap (also the riskiest period - if volume dries up, the LP doesn't earn any revenue and the whole flywheel breaks). > When the LP is high ($5M), dividends are high, so this is theoretically best time to be staking. > With sufficient volume, both states create buying/holding pressure on the token. ----- Holding/staking $PAPER is essentially betting that traders will lose against the house, which is arguably one of the most reliable bets you can make (and a good hedge if you're a degen perps trader).
proud to introduce @papertrade_xyz - a fair-launched, fully-onchain perpetuals exchange built on hyperliquid by @izebel_eth & @blurr -1000x leverage -0 slippage -No funding costs -Self-bootstrapping LP coming soon. learn more at: docs.papertrade.xyz
69
63
15
964
194,351
My favorite liang wenfeng story is that when he was in college, a buddy offered him a cofounder position in his drone startup. He turned him down to focus on academics. The drone startup became DJI. You would assume making such a decision would be the defining career moment of someone’s career. That they would regret it forever. But not this guy 😅
deepseek is raising a monster $7 billion round at $50B val making it china's largest ever AI raise but what shocks me the most is the founder, liang wenfeng: > he's personally contributing 40% of the round himself. $3 billion. > he owns 90 PERCENT of the company (unheard of at this valuation) > deepseek was founded inside his hedge fund, one of China's most successful funds. guys a fucking beast. this round is meant to achieve 2 things: 1. acquire as much compute to push out new deepseek models more often 2. turn deepseek revenue-positive by pushing new enterprise products (same tactic as OAI and anthropic) deepseek v4.1 is expected to release soon.
30
301
18
4,544
415,290
2023: Corona ended 2026: Hantavirus
28,902
88,102
58,109
397,901
Max Uper retweeted
Replying to @toly
"the cause of poverty is that smart people are creating wealth for the world" — socialists
1
3
76
5,680
Max Uper retweeted
HIP-4 mechanics decoded for short time and recurring market prediction. Last one for today. When you place a buy order on a prediction market, the system doesn't look for a seller. It creates one. Example: I buy YES at 0.40 for 100 USDH. The protocol automatically mirrors a NO order at 0.60 (because YES + NO = 1.00 always). My 100 USDH buys 250 tokens at 0.40 each. The mirror NO side: same 250 tokens at 0.60 = 150 USDH. When someone takes the NO side, a pair is minted: 1 YES + 1 NO. I get the YES, they get the NO. Total collateral locked: 100 + 150 = 250 USDH = 250 tokens x 1.00. At settlement the winner takes 1.00 per token, the loser gets 0. All collateral goes to the winning side. Same pair minting as Polymarket's Conditional Token Framework, but native to HyperCore L1. No smart contract, no ERC1155, no Polygon. Just the matching engine doing it natively in 0.07s blocks. The cheaper you buy (0.40 = you think 40% chance), the more the other side puts up (0.60). Your risk is your price. Their risk is theirs. The protocol just makes sure it always adds up to 1.00. Thank you all for the support, the feedback, and the engagement on this research. This community is something else. More digging soon. Hyperliquid
15
12
7
192
52,737
My information consumption is now 1/4 X, 1/4 podcast interviews of the smartest practitioners, 1/4 talking to the leading AI models, and 1/4 reading old books. The opportunity cost of anything else is far too high, and rising daily.
1,413
3,810
843
34,614
34,674,540
love how all the type-a adderal addicts devoted their entire lives to corporate work or finance just to be made obsolete by a clanker w/ 100 lines of code, rendering their stimulant addled corpse a useless drag on society only pot smoking creatives will thrive in a post-ai world
46
136
29
2,603
90,462
I've been dead silent about this until now: 67 days ago @alexgoughcooper asked me if I could launch his company. We went to WAR: -15 rounds of script revisions. - 2 FULL scraps to the video - 362 comments to our video editors The result? Sales calendar packed. 100+ self serve sign ups. VC's coming inbound begging to invest. We have a framework we've rinsed and repeated across 15 launches. Each of which have done MILLIONS of views and have driven Millions in pipeline. I built a 27-page document explaining the entire framework. Companies pay me $150,000 for this. Retweet + comment "Launch" and I'll send it to you.
Introducing Parker: The world's first AI Creative Director. 100+ brands like Grüns, Lume and Legends use us already. RT + Comment "Parker" and I'll send you 100+ AI prompt engineers for hire.
404
175
9
390
124,982
Max Uper retweeted
dead money theory
105
82
37
725
114,582