META BRAIN MODEL | NEW AGENT BILL | FABLE BACK SOON? nitter.app/i/broadcasts/1nxeLLOWp…

the credit card companies are launching the thing that makes credit card companies unnecessary?

SITUATION DETECTED: OpenAI engineers developed an optimization that cut inference costs in half for the models it was applied to, per The Information.

SITUATION DETECTED: Open Standard has launched Open USD, a new stablecoin backed by over 140 companies including Stripe, Visa, Mastercard, Coinbase, BlackRock, and Google.

SITUATION EXPLAINED: What happens to biology when AI can design things we don't understand and can't explain? We asked @NikoMcCarty, founding editor of @AsimovPress: "It's not that biology is irreducibly complex. Biology is governed by the laws of molecules and the laws of physics. Cells are not sacred. They're bags of molecules." "For synthetic biology to scale, for us to build very complex functions programmably inside of living cells, we probably have to get out of this reductionist historical approach where we want to understand all the genes that we are building with." "We're just gonna see much, much more use of AI to design circuits and cell functions that we don't really mechanistically understand." "There will probably be a schism where people will be using AI to design biology, and then others will try to understand why that works. You kind of try to solve the mechanism after you've already made the thing." "There will always be people who want to build things from the bottom up. I wanna understand everything I'm building with. It's just, I think that progress will be slow because it has been slow for the last twenty-five years."

SITUATION EXPLAINED: Why haven't the most intelligence-intensive fields dominated compute demand yet? We asked @braeden_norris, investor at @PropelVC: "We are still early to the higher returns on intelligence fields really capturing more of the compute supply out there. It's probably because it's just more nuanced in how it's applied. It's research-y. It's frontier. It's discovery. The capabilities are harder to manifest in the models as well." "People at the labs have talked about how Eli Lilly should be willing to spend basically infinite inference and time for a model to discover drugs for them." "Right now it's much more focused on: Eli Lilly wants to speed up their clinical trial administration process. Which is very much the same shape of work and probably not as token consumptive. And maybe can even be put on an open model because it's the same shape of work."

SITUATION EXPLAINED: What happens to AI labs' margins when the gap between open source and frontier models shrinks to a month? We asked @braeden_norris, investor at @PropelVC "The labs have had this incredible ramp in terms of revenue. But the social tide is changing. People are thinking more about costs." "They're in this constant battle between each other and also between open models, keeping that gap between near frontier and frontier alive. The existence of these margins is probably reliant on there being a gap that's somewhat substantial and isn't like a month long or even shorter." "There are plenty of use cases, drug discovery, material science, where the marginal cost you're willing to pay for a token is basically infinite. And so this defends the eighty percent plus gross margin potentially." "As an investor, that's a precarious spot to be in as a lab. It's a running race. You have to raise a lot of capital. You're investing more capital over time in R&D even if you are spinning off good margin on the inference side."

SITUATION EXPLAINED: What even is a distillation attack? @DanielleFong, founder of LightCell Energy: "I understand you don't wanna get automatically hit by a bunch of subscriptions and then distilled by DeepSeek. I do get it." "But you learned from the intellectual output of all of humanity. And we will be learning from the transcripts of what we work with with AI. This is inevitable. This is absolutely a bidirectional thing." "Is putting all my transcripts and tweets into a vector database and having the AI check every turn for something relevant... is that a distillation attack? I would argue no, but it certainly is transferring information."

SITUATION EXPLAINED: Why does every serious company need open weights now? We asked @DanielleFong, founder of LightCell Energy: "I was working with Fable on a giant live real-time radiative sim. Fable was helping with this quite a lot and then was not available right at the point where we had really settled the physics. I had to spend $5,000 worth of API spend in order to make the talk happen." "If I have a capability that these models have developed to harness and then the model goes away, if I don't have any fallback, I can't even do it." "Silent degradation, if you were detected to be working on frontier model stuff. It was in the system card for Fable that it would do this. No one liked this at all. And there are lots of times when I'm wondering, am I being silently degraded, and it's just making stuff up." "The thing about open weights in addition to being able to serve it is you're able to do modifications, understand what's happening, figure out where the model has its greatest tensions and is getting confused."

SITUATION EXPLAINED: How completely has the online/offline dynamic inverted? @Signulll, founder of @Skye: "It used to be that if you met somebody from online, it was like, 'Man, this person must be a weirdo if they're online.' And then now it's like, 'Wait, they don't have an online presence? They must be a weirdo.'" "Back in the day when people met on a dating app, they would be like, 'Oh, you know what? We're gonna tell people that we met at Whole Foods,' 'cause it was too embarrassing that you met on Tinder." "Now it's really strange if you don't have an Instagram or a Twitter account, 'cause people do this sleuthing and research on people way before even meeting them." "If I ever went on a date with Theo, I'm gonna go read his goddamn Twitter account... they're trying to gauge a picture of you based on some shitposting or whatever."

SITUATION EXPLAINED: Why is building anything new for consumers so hard right now? We asked @Signulll, founder of @Skye: "Reed Hastings from Netflix did an interview on stage and he was like, 'You know what? The only competitor to Netflix right now is kind of sleep.' Like, there's just not enough time." "The internet is now in a very mature phase. Internet acceleration probably happened somewhere around 2005, '6, '7 when everybody basically came online with Facebook. And now you're kind of in this post-teenage phase." "The supply side of the internet is growing so, so quickly, especially with the AI stuff. The cost of production is so low." "The culture shifted so drastically that online comes first and offline comes after. There's sort of this unpredictability to offline elements... you don't know what you're gonna get in terms of entertainment or dopamine hit." "Whereas online it's super predictable. If you go on Instagram Reels or TikTok, you know you're gonna have a relatively decent and predictable time. And oftentimes people are so entertained that they ended up spending like two hours inadvertently." "Everything basically competes with the predictable dopamine hit of online culture. Almost all of offline culture competes with that dynamic. And that is a very difficult dynamic from a competition perspective to win in terms of allocation of time."