🦇 This was probably my coolest cosplay of 2023 🥹 Morrigan was a challenge because I didn’t know if I could look like the character, but I really liked the result 🥰 Did you like it? 🔥 Morrigan Set available with 70% OFF now! 🔥

I'm offering Jennifer Welch and Angie Sullivan of the I've Had It Podcast an opportunity to prove themselves right. They said that homeschooling is for "dumb people" who, for some reason, want to be around their kids. So, LET’S PUT IT TO THE TEST. If homeschoolers are taught by a bunch of weird, dumb parents, it should be EASY for you to outsmart them. I challenge you to debate a 15-year-old homeschooled kid from my audience. I'll host it live on my show. It should be an easy win, right?

I've been enjoying Victoria Whitworth's new work, The Book of Kells: Unlocking the Enigma. I've actually never seen the Book of Kells in person, somewhat to my embarrassment. I've been doing some reading about the origins of Christianity this year, however, and I figured I should know something about the most famous Irish manuscript. (Perhaps the most famous manuscript, full stop.) Reading the book, I was struck by how much the contents have suffered over the past ~1200 years (enduring everything from water damage to reckless malfeasance in attempted nineteenth century restoration), and I wondered whether AI could help give a sense for how the work might originally have appeared. I downloaded the Internet Archive's PDF and asked my friendly neighborhood agent to use gpt-image-2 to render each page the way it imagines it might have originally appeared. Remarkably, this all worked with a single prompt, with the agent spinning up 48 workers, since each page took a minute or two. (I'm sure that someone wiser than me could prompt the model better, ensuring somewhat more historical accuracy in color restoration and so forth. There is no gold leaf in the Book of Kells!) This part of the project went from conception to completion before I'd finished my morning coffee. I then wanted some easy way to view the results online, so I asked Stripe Projects ( to host the result on Vercel. That also worked in basically a single prompt: I also figured that people might want an easy way to download the full PDF of updated images, but it's a large (~200MB) file, so I decided that I should charge $0.10 to cover bandwidth costs using @MPP. I asked my agent to set this up, and it basically worked smoothly, though I had to tell it what MPP is (I guess it's not yet in the pretrain) and also manually set up the Cloudflare account that actually hosts the PDF and configure the API key. (Vercel seemingly has a 100MB limit.) The purchases now show up in my Stripe account alongside all other activity. The site now has a ready-made agent prompt for anyone who wants to download the whole thing. I'm guessing that we'll see a lot more UIs like this in the future. I remain pretty intrigued by the intersection of agents, micropayments, and stablecoins. I don't know much about managing crypto wallets from the CLI, but now AI can do that for me, while Stripe seamlessly handles turning it all back into fiat. So what is the moral of the story? • Whitworth's book is very good, and you should buy it. • The Internet Archive continues to be wonderful and a civilizational treasure. • While there are rough edges, setting up third-party services via the CLI now basically works. I'm pretty sure I wouldn't have bothered with any of this if I couldn't have outsourced almost all of the work to AI. • The image models have gotten very good. • There will probably continue to be all kinds of interesting applications of AI to history. (The Vesuvius Challenge of course being a shining pioneer.) • These days, I often find myself building single-use sites for things I'm learning or for books I'm reading. I think this is a cool new category of software.

Making AI "reason about space in words" might be backfiring 🧭 Here's a new approach that lets it imagine unseen viewpoints instead. Title: Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models URL: 🧭 Overview This work proposes Imaginative Perception Tokens (IPT) to strengthen spatial reasoning in vision language models (VLMs). Rather than forcing spatial logic through language, it keeps "what could be perceived under a different arrangement" as an intermediate perceptual representation. ❓ Challenges Solved VLMs struggle with spatial reasoning: inferring unobserved viewpoints, reasoning through occluded paths, and integrating partial observations. Prior work pushed this into textual chain-of-thought, but forcing visual reasoning through language alone hit a ceiling. 💡 Methodology & Proposed Approach ・Uses the unified VLM backbone BAGEL, trained with IPT supervision ・Formulates three tasks: Perspective Taking (PET), Path Tracing (PT), Multiview Counting (MVC) ・Builds a ~20,000-example dataset with ground truth, answers, and metrics The core idea is treating the perception itself ("if I moved here, I'd see this") as an intermediate representation. 📊 Experimental Results ・IPT improved Multiview Counting (MVC) accuracy by 3.4% ・Path Tracing (PT) reached performance competitive with closed-source models ・IPT supervision outperformed textual chain-of-thought training ・Conversely, textual CoT substantially degraded spatial reasoning #SpatialReasoning# #MultimodalLLM#

Beat Writer @DarnellWrites joined us on @espnthefan to discuss BYU Basketball's nonconference schedule and the matchup fans are most excited about: Q: What was your first reaction to BYU’s nonconference schedule? Dickson: My first reaction was that it's very different from last year's schedule. With AJ Dybantsa on the roster, BYU had an opportunity to showcase him on some of college basketball's biggest stages. That helped open doors to games at TD Garden, Madison Square Garden against UConn, and Las Vegas. It was one of the most high-profile nonconference schedules BYU has ever played. This year's schedule isn't as flashy, but it accomplishes what it needs to. BYU is going to face one of the toughest conference schedules in the country in the Big 12, so there's really no reason to overload the nonconference slate. There are still plenty of quality opportunities. I like the exhibition schedule with Nebraska and UNLV. Opening the regular season against Ohio State at the Delta Center is a strong matchup, and the Maui Invitational gives them another chance to play quality competition. One thing I also really like is seeing Utah State and Weber State back on the schedule. For the first couple of years under Kevin Young, BYU didn't play any in-state opponents outside of Utah, and that was the first time since the early 1960s that had happened. Utah State is a very good team, and it's always fun to see those local rivalries return. Ultimately, this schedule is about preparing for the grind of the Big 12, which we all know is a gauntlet. Q: Fans seem especially excited about the Utah State and Weber State games. Why do you think those stand out? Dickson: I think people enjoy the in-state matchups. Mark Pope used to call those games "magic," and while I don't know if I'd go that far, they definitely create a different atmosphere. Those games matter to fans. When you look at the rest of the home nonconference schedule—Queens, Omaha, Montana State, Cal State Fullerton, La Salle—those aren't games that move the needle very much. But Utah State and Weber State do. They're local programs, there's familiarity, and the Marriott Center should have a great atmosphere for both. I remember talking with Chris Burgess last year about scheduling philosophy, and he mentioned looking at Arizona's model. Arizona challenged itself in the nonconference, even taking some losses, but those quality games helped its NCAA Tournament résumé. It's always a balancing act. This year's team is still adding new pieces and building chemistry, so you don't necessarily want to overload the schedule before conference play begins. Q: Looking at the schedule today, what's your projection for BYU's nonconference record? Dickson: I think anything more than two losses would be a real disappointment. The schedule sets up well. Most of the games are at home, with two at the Delta Center. The only significant travel comes with the Maui Invitational and the road game at San Diego State. We still don't know exactly what the final roster will look like because BYU still has scholarships available and is looking to add another big man. But based on what we know today, I think this team has a chance to run the table. At worst, they should be around one or two losses heading into Big 12 play. If they lose more than two games in the nonconference, I think that would be disappointing given the opportunities they have with this schedule.

Cost per performance in AI seems to be falling more than ~200x annualized (and higher at higher levels of benchmark performance.) And exceeding 500x annualized cost declines moving along the efficient frontier of the performance curve. Probably more relevant practically for users, at $1 per task whereas today you might only have 50/50 odds that the agent succeeds, by the end of year, you should have an 89% chance of success and by a year from now a 97% chance (at least on deepSWE 1.1 type tasks.) The moving pareto frontier makes it difficult to cleanly report a performance cost improvement due to the shape of the cost performance curve. At the highest asymptote of performance you go from literally not being able to achieve such a low error rate *at any cost* to being able to get that performance for 10s of $s per task (an infinite cost decline). We forecast that top-end rate separately. In the belly of the curve you need only cross a performance cost threshold, but as the curve steepens you can also less expensively buy more performance (though not 100% clear that the curve really is steepening that much.) We also model that performance curve, though given uncertainty effectively zero out continued improvement for the conservative case. And at lower levels of benchmark performance you get a cleaner understanding of the underlying cost-per-performance improvement though at thresholds that are less interesting practically. If anything I suspect that people are wildly under indexing on the rate of improvement of these models. Even last month's AI spend will fall to a fraction of a fraction within the year (if we weren't going to so predictably deploy the new capabilities against new workloads, tasks and challenges.) Spending a lot of time trying to painstakingly carve out specific current workflows into more efficient models, at this point in the improvement curve, seems, if anything, a fool's errand designed to line the pockets of a consultant. Let the tokens flow.

To those who say "it's her choice" I used to say the same thing about my own Hijab. I was covered head to toe. Even on the beach. Even with my male relatives. And when people challenged me, I said: this is my choice. You know why? Because the other option was burning in hell, hung by my hair for eternity. That's what they taught me since childhood. Yes, let me repeat it for you to just understand about our version of “freedom of choice.” I grew up in an Islamist culture that gave me two options. Cover yourself and go to heaven. Or don't cover, and be hanged by your hair and burned in hell. Like when someone puts a gun to your head and says "you can go left or right", technically, yes, that's a choice. I picked heaven. I called it freedom. And I believed it, because that's what happens when fear is all you've ever known. And if you think I shouldn't be telling you my story, I don't care. I really, genuinely, from the bottom of my heart, don't care. The women still living under that "choice"? I care about them. #LetUsTalk#

Peter Diamandis in a new interview on CNBC this morning said @SpaceX has the potential to become the world’s first $10 trillion company. "Well, if anybody can do it, Elon can. I think this has the potential to really be the first $10 trillion company out there and growing. He’s building civilizational-level infrastructure. The challenge is I don’t think the market knows how to value this company. It’s not a company. It’s really four different businesses that are wrapped into a vertical, and you have to look at each individually. The first business they’re in is global communications with Starlink. Starlink is going to be delivering global gigabit connectivity to the entire planet, billions of individuals. They’re just making a dent in that regard right now. And then they’re going direct satellite-to-cell phone. As Gwynne said on the earnings call yesterday, that’s a $600 billion level of revenue potential, and they expect to take a chunk of that. If you can be driving anywhere on the planet, in the middle of the mountains or in the middle of the outback, and get cell phone connectivity, that’s the service I’m going to want to use on a global basis. So the telecom industry on its own has the potential to deliver a large chunk of that. The second business that they’re in is as a hyperscaler delivering AI. They’ve already signed up $28 billion of annualized ARR in terms of revenue for that business, and they’re building out faster than anybody else is doing it. They announced yesterday they’ll be at 2 GW of compute at the end of this year and shooting for upwards of 10 GW by next year. And that business alone is where they’re going to be having probably the greatest amount of revenue, especially as they go beyond that to building orbital compute. Surprisingly, Elon said they’re going to be launching their first AI satellites in partnership with Nvidia in 2027, much faster than anybody thought they would. And if they hit their numbers, we’re talking about an infrastructure of AI compute in space that’s going to be serving all of the hyperscalers, because no one’s going to be building that fast."