0:00 California just officially closed the “Montana license plate loophole” for good.0:05 That means the trick of using a shell company in a no-sales-tax state to register your supercar and dodge California taxes is finally over.0:14 And while last week we were talking about AI forging cartoons, this week the AI story got a whole lot more...0:21 formal, but we'll get to that.0:23 First, let's talk about why this tax law change is a bigger deal than it sounds.0:29 For years, if you were wealthy enough, you could hire a firm in Montana to create a shell LLC for you.0:35 That LLC would "own" your Ferrari or your half-million-dollar R.V., and you'd drive it around Los Angeles with a Montana plate, having paid zero California sales tax.0:46 The old law was flimsy — it only required that the company be fifty percent owned by out-of-state residents, a test that was trivial to pass.0:56 But the new bill, SB 1406, which passed on September thirtieth, completely rewrites the definition of a shell company.1:04 It doesn't matter who owns it anymore.1:06 The new rule is simple: if a California resident is a beneficial owner or officer of the company, they are now personally on the hook for all the unpaid taxes, plus interest and penalties.1:19 It's a surgical strike against a very specific, very expensive form of tax evasion.1:24 Now, for the rest of the week's biggest threads.1:27 The top story, generating a massive discussion, is that OpenAI just published a major update on its progress in mathematics.1:36 They released a whole repository of AI-generated mathematical preprints and, crucially, formalized proofs written in a language called Lean.1:45 This means the proofs can be verified by a computer, which is a huge step for trusting an AI's output.1:52 They're even sharing the compute cost: about three hours of ChatGPT Pro usage per result.1:58 It's a surprisingly transparent move.2:00 And in a one-two punch, OpenAI also announced the public beta of a new "Decisions API" today, October seventh.2:07 This isn't just about generating text; it's designed to help developers build workflows where the AI makes structured choices.2:16 Think of it as moving from a creative partner to a logical one.2:20 The community is still digging in, but it points to a future where you ask an AI not just to write something, but to decide something.2:29 Then there’s a story that has the gaming and emulation scene buzzing.2:33 A project called AnyPS5 just reported a massive milestone.2:37 They've now successfully mapped eighty-seven percent of the PlayStation 5's system libraries.2:44 Their goal is to run PS5 games natively on a PC, without the performance hit of traditional emulation.2:50 This isn't about playing pirated games; it's a monumental reverse-engineering challenge.2:56 Getting to eighty-seven percent means they are getting seriously close to making this a reality.3:03 It's one of those deeply technical, multi-year projects that you just have to admire.3:08 On the flip side of a massive undertaking, we have a story of a single developer's frustration.3:14 The creator of Photopea, that incredibly popular browser-based photo editor, posted that he’s been fighting with GitHub for a month.3:23 He filed DMCA takedown notices for cracked, pirated copies of his software being hosted on the platform, and GitHub refused to take them down.3:33 Their reason?3:34 They couldn't confirm it violated a very specific part of the law.3:38 The developer's post is full of raw frustration, and he suspects his reports were never even seen by a human.3:45 It's a classic David-and-Goliath story for the platform age.3:49 And finally, a new open-source project caught a lot of eyes.3:53 It's an email client for Linux called Penguin Mail, written in Rust.3:58 It aims to be a modern, all-in-one inbox, pulling in Gmail, Microsoft accounts, and standard IMAP, plus calendar and contact integration.4:07 It also has built-in OpenPGP encryption and an optional AI assistant that can run locally on your machine.4:14 In an era where everyone is trying to escape the browser, a slick, native email client is always going to get attention, especially one built with privacy in mind.4:25 Okay, let's go deeper on two of these, because they represent two really different, but equally important, struggles happening in tech right now.4:34 The first is about creating new knowledge, and the second is about protecting what you've already created.4:41 Let's start with OpenAI and the math proofs.4:44 This is so much more than just "AI is good at math now." The key is in the details.4:50 They aren't just spitting out answers.4:52 They're providing formalized proofs in a language called Lean.4:56 So what does that actually mean?4:59 Think of a normal mathematical proof written in English.5:02 It relies on a shared understanding between mathematicians.5:06 It has logical steps, sure, but there's a bit of...5:10 interpretive dance.5:11 A human has to read it and agree that it makes sense.5:14 A formalized proof in Lean is different.5:17 It's written in a programming language where every single logical step is so precise, so unambiguous, that a computer program can check it for correctness from the axioms up.5:29 There's no room for error.5:30 If the Lean proof "compiles," the math is correct.5:34 Period.5:34 This is where we've seen this pattern before, but with a twist.5:38 The classic example is the Four Color Theorem, which states that any map can be colored with just four colors without any adjacent regions sharing a color.5:49 It was first proven in 1976 with the help of a computer, which checked thousands of different cases.5:55 It was controversial.5:57 Mathematicians complained they couldn't verify it by hand; it was too complex.6:02 They had to trust the computer's brute-force checking.6:05 But what OpenAI is doing is different.6:08 This isn't just a computer checking a human's work.6:11 This is an AI generating novel mathematical ideas and then providing the machine-checkable proof itself.6:18 It’s like the computer isn't just the calculator; it's the mathematician.6:23 And by releasing the proofs in Lean, they're handing the community the ultimate tool for verification.6:30 You don't have to trust OpenAI.6:32 You don't have to trust the black box of the neural network.6:36 You just have to trust the Lean compiler, which is open source and heavily scrutinized.6:42 This feels like a direct response to the credibility crisis in AI.6:46 How do you trust the output?6:48 Well, in math, you provide a proof.6:50 And by also publishing the compute cost — "three hours of ChatGPT Pro" — they're demystifying the process.6:57 It's not some unobtainable magic from a trillion-dollar datacenter.7:02 It's a quantifiable amount of work.7:04 They're saying, "Here is our result, here is the verifiable proof, and here is what it cost us to find it." It's an incredible flex.7:13 This also changes the job of a human mathematician.7:16 It doesn't make them obsolete.7:18 Instead, it turns them into explorers and guides.7:22 The AI can generate hundreds of potential paths or interesting conjectures, and the human's job is to provide the intuition, to ask the right questions, to point the AI in a fruitful direction.7:34 It's a collaboration.7:36 And the fact that OpenAI is working with the Institute for Advanced Study—where Einstein used to work—shows they understand this.7:44 They're not just dropping code; they're trying to build a bridge to the existing world of pure mathematics.7:52 This is how you get a deeply conservative field like math to actually engage with AI.7:57 You don't just give them answers.7:59 You give them tools and proofs.8:01 Now, let's switch gears completely.8:04 From the frontier of knowledge to the grim reality of running a business online.8:09 Let's talk about Photopea and GitHub.8:12 The developer, who goes by IvanK_net, built something amazing.8:16 Photopea is a free, powerful photo editor that runs entirely in your browser.8:21 It's a lifeline for students, schools, and people who can't afford a Photoshop subscription.8:27 He makes money from ads and a premium version.8:30 But because it's written in JavaScript, people can just copy the code, re-host it, and sometimes inject their own ads or malware.8:39 So Ivan does what you're supposed to do.8:41 He files a DMCA takedown notice with GitHub, where these cracked copies are being hosted.8:47 A month later, he gets a rejection.8:49 And the reason GitHub gives is incredibly specific and, frankly, infuriating.8:55 They said they couldn't confirm a violation of 17 U.S.8:58 Code § 1201.8:59 So what does that even mean?9:01 Section 1201 of the copyright act isn't about copying the work itself.9:05 It's about circumventing "technical protection measures." Think cracking the copy protection on a DVD or a video game.9:13 GitHub's response implies that because Photopea's code is just plain JavaScript, and there's no complex encryption to "crack," simply copying it doesn't violate that specific part of the law.9:26 This is the kind of hyper-literal, legalistic response that makes you want to tear your hair out.9:32 It completely misses the point.9:34 The code is still copyrighted.9:36 Hosting an unauthorized copy is still infringement.9:40 But GitHub seems to be hiding behind this narrow, almost procedural excuse.9:45 As the developer said, "I really doubt that a real person ever looked at my report." And he's probably right.9:52 Here's the pattern: this is the curse of scale.9:55 We've seen this with YouTube's Content ID, with Amazon's counterfeit problem, with social media's content moderation.10:03 When you operate a platform with billions of pieces of user-generated content, you can't use humans to review everything.10:11 You HAVE to rely on automation.10:13 And that automation is optimized to protect the platform from legal liability, not to deliver justice for individual creators.10:22 GitHub's automated system likely saw the report, ran a check against a narrow set of criteria for the most clear-cut, slam-dunk violations, and because this case involved JavaScript code and not a cracked executable, the algorithm just said "nope" and sent the rejection template.10:40 The system is designed to say no.10:42 A false negative—leaving infringing content up—is safer for the platform than a false positive—taking down legitimate content and getting sued.10:52 So the creator is left in an impossible position.10:55 He's up against a system that isn't just broken, but is working exactly as its owner designed it to—to minimize its own risk, even at the expense of the users it claims to serve.11:07 The comments on Hacker News were full of sympathy, but also a sense of resignation.11:12 Some people suggested porting the app to WebAssembly to make it harder to copy.11:18 Others just said, yeah, piracy is a fact of life.11:21 But that misses the core issue.11:23 This isn't just about piracy.11:25 It's about a breakdown of the social contract of platforms.11:29 The promise of a platform like GitHub is that it provides tools and infrastructure so you can focus on building.11:36 But when that same platform becomes a safe haven for people stealing your work, and its own enforcement mechanism is a black box that spits out nonsensical rejections, the promise is broken.11:49 It shows that for all the talk of "community," when push comes to shove, the platform's automated legal defense system will always win against a single human.11:59 So what does it all add up to?12:01 On one hand, you have AI pushing the absolute boundaries of what we can know, creating new systems of logic that are verifiable and transparent.12:11 And on the other, you have our existing human systems—our legal and platform systems—failing in the most basic ways, becoming opaque and unaccountable.12:21 It's a perfect snapshot of the weird moment we're in.12:25 We're building these god-like tools for discovering truth, while the everyday systems we rely on for fairness feel like they're falling apart.12:34 This week sets up a fascinating tension.12:37 We're seeing the birth of AI as a genuine partner in formal science, a tool that doesn't just guess but proves.12:44 The next six months aren't going to be about whether AI can do math, but about how human institutions adapt to it.12:52 Will universities create new courses?12:54 Will journals accept AI-co-authored papers?12:57 That's the real test.12:58 At the same time, the struggle of individual creators against massive, automated platforms is only getting more acute.13:06 The real story to watch isn't whether GitHub changes its policy for one developer, but whether the growing frustration forces a broader reckoning with platform responsibility.13:18 We are simultaneously building systems of breathtaking power and failing to maintain the simple ones we already have.13:26 And the friction between those two is where everything important is going to happen next.
