Pacing The Frontier is both (imo) the correct thing to do, but also incredibly beneficial to the frontier AI labs from an economics perspective. It will allow them to slowly roll out intelligence, amortize models over longer periods of time, experience margin expansion over a lifecycle of a model for longer, and embed existing models into workflows that the longer they are "frontier" the lower churn will be (it will still be high). It de-escalates the prisoner's dilemma and also allows them and the whole industry to exhale a bit on 2030+ capex existentialism even if just for a year or two, which matters a lot. All of these things then flow into the idea that safety/alignment could have more resources, both financial (more FCF for the labs to deploy), as well as compute (inference scales perhaps less aggressively, post-training holds steady, pre-training maybe re-allocates a bit, and social capital wise safety/alignment viewed as main blocker to restart the other machines and thus becomes higher priority). The main areas where this all is negative is that likely the neolabs would have less constraints placed upon them due to their (lack of) compute scale, and of course the whole "what do we do with china" problem. I am in agreement with some's POV (@hamandcheese notably discusses this as do others found in @TheZvi latest round up - that we figure that out later though, and I hold some strange belief that outside of this, I could see China actually getting to a point of Pacing on their own. I understand that many overintellectualize the incentives of the labs, and I'm not saying that Pacing is directly coming from an economic incentive versus a "long-tail let's not destroy humanity" one, but I do think it is hard to ignore that if regulatory capture in the traditional form doesn't work, Pacing is so economically dominant, and is a rare occasion to let air out of a bubble that is driven by a feeling of existential competition.

@j_asminewang Hi Jasmine. How many of your OpenAI shares will you bet me that humans are not extinct in 2030? You can probability weight your wager. That will tell me everything I need to know. I am good for any amount you choose. I don’t mind pacing - I hate BS scare tactics. 🧐

We wanted something that felt uniquely “Compound” & also added gravity to the moment that is choosing to partner with us for the next (hopefully) decade. We worked with a Brooklyn studio @cwandt to customize their single-use launch clock, Time Since Launch.

We just confirmed two new speakers! Stanford researcher on her mechanistic work on brain circuits that drive sleep regulation And UCSF epidemiologist on her evidence for the link between sleep & AD/PD You won't find more wisdom about sleep in one room

We crossed $200M ARR last week, doubling from $100M in April. Runway is the world model company. We power the daily work of more than 60 million creators and 95% of the Fortune 100. From award-winning films and ad campaigns to real-time media applications to generalist robotics. Our largest customers are increasingly licensing our frontier models directly. As we expand our research in robotics, the wonderful team from Kinetix, a Paris-based research lab, is joining Runway to help drive our efforts there. Proud of the team for getting us here. Back to building the universal world simulator.

Long running scientific tasks benefit hugely from external structuring, and graphs are a great way to do that. CurieOS automatically structures tasks with DAGs behind the scenes for coherence in the hardest scientific and engineering problems

Data sales is one of the most important areas in AI today. It matters both for improved model performance, realizing true AI world models and especially for deeptech startups that now have a new monetization path. However, the best way to build companies and durable revenue streams around data sales in deeptech areas is largely unsolved outside of Mercor and Scale. We @CompoundVC have a very clear thesis for what platform <> data sales alignment looks like in biotech. Namely that companies shouldn't go significantly off their core path to sell data to frontier labs. We also believe that only biologically complex and general platforms make sense as data providers in the long term. This thesis is only one part of the puzzle. There's a whole world of materials, energy, health, and robotics that are still figuring out durable sales too that we want to discuss in depth. At this Compound Research Day on Data Factories we will have several deeptech companies (including in bio!) showcase their data generation platforms, specifically focusing on data novelty and scale for model training. Operators at frontier and neolabs will also be in attendance and speak. Join us to discuss one of the most consequential areas in AI! (We're still building out the speaker list so let us know if you'd like to present.)

I returned to the cafe that gave me caffeinated coffee instead of decaf to conduct some investigative work. Here is what we know. I entered the cafe on Sunday, Aug 30th at 2:06 pm. I took roughly 15 steps to the ordering station (see image below). A woman in her early 20s cheerfully greeted me. “How may I help you?” she inquired. “A decaf coffee please” I said, matching her lightness. “Do you want any milk?” she delivered with mastery. “No thank you” I said delicately to avoid offending her generous offer. “A name for the order?” “Bryan” “Ok, that will be $5.09” I then double clicked the right button on my phone and tapped the bluetooth device. Then, unexpectedly, this woman smiled, looked at the device I was holding in my other hand, and said “I like your walkie talkie”. This took me by surprise. She had deviated from the script and routine. What was I to make of this? She then flipped the payment screen around for signature and simultaneously turned away, as if to allow me to do my business in private. I scribbled nonsense onto the screen and was presented with a tip conundrum of $1, $2, and “Other”. What was I to do? She had clear mastery of her craft and had provided great service thus far. But I was suffering a debilitating aftershock from the surprise compliment and friendly engagement. $1 would be shameful. $2 seemed more appropriate. That would be a 39% tip for the sixty second encounter. But I am rich. Should I give more? Should I give all? An existential dilemma draped over me. My fingers dashed to touch $2 and end the spiral. I then walked to the opposite side of the cafe where the barista would be fulfilling the order and handed them my stainless steel cup. This was to avoid using the microplastic lined material they would have otherwise used. I’d brought with me a flask of warmed, freshly made macadamia nut milk. Before walking to the cafe, I had heated 8 oz of nut milk to 145°F, near the ceiling before plant proteins denature and aggregate and the emulsion destabilizes. The ten minute walk inside my insulated flask would keep it around 140°F and pair perfectly with the 185°F freshly brewed coffee. This would be my companion over the next few hours on this splendid Sunday afternoon. As I headed to leave the cafe, I ran into some friends and engaged in a lively conversation for four minutes. I wrapped the conversation and left the building for the coffee pick up location outside. My cup and the coffee were on a ledge with a paper order receipt below it. I picked up the cup and eyed the BPA-coated thermal paper receipt that unfortunately needed handling lest I litter. Such a moral dilemma. In this ethical deliberation, I forgot to check the receipt confirmed “decaf.” Rookie mistake. I walked home with the anticipatory delight of savoring the first touch with my lips. I’d wait until at my desk, fingers on the keyboard, music in the ears and upright posture. Then the sipping began. It was glorious. At 5:43 pm I messaged Kate “the nap worked, I’m still going strong”. Earlier in the day I’d told Kate my plan was to complete 10 hours of deep, focused work. It was Sunday, my calendar was open and it was a unique opportunity to work in an interruption free stretch. To do this, my plan was to take a mid-day nap to refresh myself. So earlier that day, at 12:12 pm, I’d taken a 31 minute nap with 94% efficiency. I woke up feeling refreshed. Recently I’d been dipping with fatigue in the 2-4 pm window. Likely because I’ve been waking up at 4 am and completing an intense five hour morning routine (exercise, sauna, IHHT/HBOT) and then working 10 hours straight, until 7 pm. The nap did work. Turns out, I also made the mistake of attributing my caffeine high to my nap. Fatal. I finished work and began my wind down routine. I laid down in bed at 8:01 pm and only then did I realize I had a problem. Whereas normally I’d fall asleep within 2-3 minutes, something was off. My mind was wired and buzzing. There was no way I was going to fall asleep. Not having had caffeine in years, I tossed and turned all night, creating a subjective feeling that I was awake all night. Was this a mistake, a prank or a crime? Noise > yesterday, I returned to the cafe with a sound meter > the cafe’s background noise sat around 75 dBA > I replicated my order voice volume. It was 72 dBA > The person behind the counter is 3 feet in distance > this created a -3 dB signal to noise ratio > you can hear low-frequency vowel volume > but high-frequency consonant clarity drops off > even with healthy hearing between 1,000-4,000 Hz > “a decaf coffee” and “a cup of coffee” could sound nearly identical Ordering mistakes > I took a picture of the order screen (see image above) > that they use to input the order > the caffeinated/ decaffeinated buttons are adjacent > buttons are large enough a finger mishap is unlikely > the failure point here is a “capture error” > both buttons say “brewed” > therefore size/color do nothing to offset autopilot error > order error rate in a cafe is around 1 in every 8 orders > afternoon mistakes (when I was there) most common > there is also a possibility that the barista making the coffee messed up the order, or pranked me Recognizability and prank > given my social media presence, there is a ~55% likelihood that she recognized me > she could have thought "lol, this is going to be funny" Given this, what do you think happened?

Everything begins inside your brain. Every thought. Every decision. Every experience. And yet the organ responsible for all of it, the physical instrument shaping how you think, react, connect and recover, has remained invisible for as long as humans have existed. Today, that changes. The Atlas 1.0 is here. Pre-order now.

Huge news. President Trump appears to have confirmed our August 11th post regarding potentially imminent capital gains tax cuts. We may soon see the first major federal capital gains tax cut since 2003, when the top long-term capital gains tax rate was reduced from 20% to 15%. More to come on @KobeissiLetter.

Ben Horowitz on the oldest law in startups, and why the industry no longer has to obey it: "The one thing we all knew in startup world is that if I have a two-year lead on you and you try to catch me by hiring 1,000 engineers, you're gonna wreck your company." "That never works. It's a Mythical Man-Month. Nine women can't have a baby in a month. That's it. That never works." "Okay, now that works. But it's not hiring 1,000 engineers. It's taking $3 billion and lighting up a magnificent cluster, and then all of a sudden, whatever, Grok can come out of nowhere and, oh, all of a sudden it's real." "You can throw money at the problem, and you can throw money at almost any problem, and that works. That is just completely different than anything we've ever lived through. By the way, we're all psychologically adjusting to this." @bhorowitz @eriktorenberg

Coinbase CEO Brian Armstrong says donating to charity is net negative "The common view is that philanthropy is this noble cause, and Bill Gates is like 'I'm going to go rehabilitate my image and do philanthropy.'" "A lot of charities and philanthropies are actually net negative for the world and they get captured. It's remarkably hard to find a foundation that has not gotten captured by ideology.” “By trying to do something good, like criminal justice reform, it turns out they just increased crime. I have a donor advised fund and I put money toward things helping civilization advance, but I don't think at this moment I would ever create a foundation." "If you look at the Ford Foundation, I think Henry Ford would be turning over in his grave if he knew what they were up to.” “I have a DAF, a donor advised fund, and I just try to put money toward things that I think are helping civilization advance. But I don't think at this moment I would ever create a foundation."