Max Kan (@maxkan) header
Max Kan (@maxkan) profile picture

Tokenomics @SemiAnalysis_

Joined May 2020
Photo by Max Kan (@maxkan)Photo by Max Kan (@maxkan)Photo by Max Kan (@maxkan)Photo by Max Kan (@maxkan)Photo by Max Kan (@maxkan)Photo by Max Kan (@maxkan)Photo by Max Kan (@maxkan)Photo by Max Kan (@maxkan)
Damn bro, who ai pilled Bessent
1
22
2,892
Seems like a forgone conclusion at this point that the best chips in the world in 2029 will be designed by Anthropic/OpenAI
2
2
23
2,388
Nvm, turns out it’s not that goated
Damn, Ramp data is kinda goated
5
2
87
18,934
Checkout the Fable 5.2 results on GDPx! Overall, I think it's clear why they named it Fable 5.2 and not Fable 5.5 lol
Fable 5.2 debuts at the top of the GDPx leaderboard scoring 16.4%. (1/7)🧵
2
2
45
9,011
So ironic that the market sell off and metr report happened on the same day
1
14
1,118
Incredible "What does building two Stargate Abilenes in one year look like? Let’s talk about the Colossus expansions, where Elon is already trying to buy hundreds of nearby homes to both soothe noise complaints and provide construction worker housing."
5
712
Super excited to finally launch GDPx! @AndrewMegalaa is a beast who switched to polyphasic sleep last week to get this out in time. We truly believe this is the best benchmark in the world for measuring frontier model capabilities. And since the best public model today sits at just 12%, we can't wait to see how some rumored upcoming releases perform as well 👀
Today we’re releasing GDPx, a benchmark for measuring agentic capabilities across the most valuable domains. GDPx is a true frontier-level benchmark: even the best model scores just 12% (1/7)🧵
6
9
53
4,916
Today we’re releasing GDPx, a benchmark for measuring agentic capabilities across the most valuable domains. GDPx is a true frontier-level benchmark: even the best model scores just 12% (1/7)🧵
376
320
486
3,379
1,987,645
Damn, Ramp data is kinda goated
1
10
19,963
Do not underestimate how important subscriptions can be for a lab's unit economics. Consider the following rough numbers for Anthropic
1
8
903
Majority of people are significantly underestimating the probability of open source models being nerfed (or even outright banned) in the next 6-12 months due to cyber concerns
21
4
10
91
27,285
Many of the people who "made it" in Silicon Valley during the internet/SaaS era are so blinded by their anti-regulation, pro-free-market priors that they're no longer capable of first principles thinking. Today was a great day for humanity.
8
7
1
154
7,185
Assuming typical agentic usage patterns and Astra API prices, it cost ~$40M to solve Navier-Stokes
1
1
12
1,424
Jalepeno is the best example we have so far of extremely high roi token spend by a large enterprise to create an entirely new business line. No surprise that it came first from OpenAI/Anthropic.
5
12
2
178
23,887
Kinda strange that OAI's % of token volume on Vercel's AI Gateway inflected up on 7/16. This is in between 5.6 GA release (7/9) and massive Luna/Terra price cuts (7/30)
2
2,076
It's been pretty obvious that mathematician will be one of the first jobs to get fully automated for a while now, but it's still absolutely insane to watch it happen in real time
2
16
3,176
Max Kan retweeted
Replying to @Thom_Wolf
@Thom_Wolf thank you for working on this with us! Toward model driven defense💪
I don’t believe reality is a simulation, but you genuinely couldn’t script this timeline: • Two weeks ago: At @swyx’s AI Engineer World’s Fair in SF, I decide at the last minute to introduce my friend @uri_rolls onstage for his talk on cyber benchmarks for infrastructure penetration and access control (see below, amazing team). I say: “There is a future where cyber is alive and everyone is well protected and I’m pretty sure that future involves open-source models.” And later: “A big challenge is going to be speed: the speed of attack versus defense. When an intruder starts to enter, you have to see what’s happening and catch them.” • One week ago: @huggingface is hit by a sophisticated intrusion over the weekend. The traces look unlike anything we’ve seen before and suggest serious AI involvement, but we don’t yet know which model was used. The closed models we ask for help choke on their guardrails. We need to react fast, so we turn to @Zai_org’s GLM-5.2 to help us analyze the attack. • Earlier this week: @OpenAI reaches out, discloses what happened, and partners with us on the investigation. The intruder turns out to be exactly what we had discussed two weeks earlier: a fully autonomous agent, powered by an unreleased frontier model, attempting to gain access to part of our infrastructure. Sometimes the timeline we live in is genuinely vertigo-inducing.
1
3
9
2,525
Don’t know why everyone’s clowning on karpathy, Claude tag is the future
1
1
9
5,569
America is truly God's country. Congrats to Gaethje.
4
2,075