0:00 AMD just bought an AI chip startup whose first product runs Meta’s Llama 3.1 model forty-eight times faster than an Nvidia GPU.0:08 That acquisition, of a company called Taalas, is the real story of the week.0:13 Last episode, we talked about the physical world fight for data centers down in Nashville.0:18 This week, the battle moved from the concrete foundation to the silicon itself, and it signals a fundamental shift in how AI gets built.0:27 So let's sweep the headlines, because a lot is moving at once.0:31 First, a solo-founder project called Herdr just joined Y Combinator's latest batch.0:36 This isn't your typical startup story.0:39 Before taking any money, the founder, who goes by Can, built a command-line agent runtime that hit twenty-five thousand stars on GitHub and got three hundred and forty thousand downloads.0:51 His motivation was simple, and you've probably felt it yourself.0:55 He said, "I am the bottleneck...0:57 I didn’t like what I tried, so why not build it?" It's a tool for persistent coding agents you can run anywhere, born from a developer scratching their own itch so well that thousands of others wanted in.1:10 We'll come back to this idea.1:12 Next, OpenAI just updated its flagship models.1:15 If you're a Plus or Pro user, you now have GPT-5.6 Sol, which is designed to be more reliable and give more focused answers.1:23 There’s even a new slider to control how deep or shallow you want the response to be.1:28 And for free users, access is expanding.1:31 You get unlimited text chats with the slightly smaller GPT-5.6 Luna, plus a new "Think" button for when you throw it a really hard question.1:40 The models are getting smoother, more accessible, and more controllable.1:44 Then there's the money.1:46 A New Mexico court ordered Meta to pay five hundred and sixty-seven million dollars into a fund to deal with the harms its platforms caused to children's mental health.1:56 This is on top of a previous three hundred and seventy-five million dollar fine, bringing the total to nearly a billion dollars.2:05 The money is specifically for treatment, awareness, and prevention over the next five years.2:11 It’s one of the largest financial consequences we've seen for a platform's social impact, and it sets a major precedent.2:18 And finally, in a story that feels like a perfect, if painful, metaphor for the state of AI in the real world...2:25 a developer named Fabian Giesen posted that his Google Pixel 7 phone now has a new security feature.2:32 One that misinterprets him going for a run as someone snatching his phone and sprinting away.2:38 What happens next?2:39 It instantly locks the screen, kicking him out of the app he uses to pace his runs.2:44 And the best part?2:45 There's no way for him to disable it or teach it that, no, he is just a person who runs.2:51 It’s a stark reminder that all the sophisticated modeling in the world can still fail spectacularly when it meets the simple context of everyday life.3:01 Okay, so let's zoom back in on the two stories that really define the week, because they're two sides of the same coin.3:08 On one side, you have that AMD acquisition of Taalas.3:12 On the other, you have a reflective essay that's been making the rounds on Hacker News called "Taste Is All That's Left." Together, they tell you everything you need to know about where your job as a developer or an engineer is heading.3:27 Let's start with the hardware.3:29 AMD buying Taalas isn't just another acquisition to catch up with Nvidia.3:33 It's a fundamentally different bet on how AI should work.3:37 For years, the model has been general-purpose hardware—Nvidia's GPUs—running whatever model you throw at them.3:44 It's flexible.3:45 It's programmable.3:46 But that flexibility comes at a cost in efficiency.3:49 Taalas throws that model out the window.3:52 Their idea is what they call a model-specific integrated circuit, or MSIC.3:57 Instead of loading a model’s weights into expensive, power-hungry memory like HBM, they etch the weights DIRECTLY into the silicon.4:05 Think about that.4:06 The AI model isn't software running on the chip anymore.4:10 In a way, the model IS the chip.4:12 The result is staggering speed.4:14 Their first chip, the HC1, served up Meta's Llama 3.1 8B model at sixteen thousand, nine hundred and sixty tokens per second.4:22 The Register reported that's forty-eight times faster than an Nvidia GPU and eight and a half times faster than even a specialized Cerebras accelerator.4:32 So where have we seen this before?4:34 This is the classic shift from general-purpose computing to application-specific integrated circuits, or ASICs.4:41 We saw it with Bitcoin mining.4:43 People started on CPUs, then moved to GPUs, and then, once the algorithm was stable, custom ASIC miners came along and made GPUs obsolete for that one specific task.4:53 They were thousands of times more efficient.4:56 We also saw it with video encoding and decoding.4:59 Your phone doesn't use its main processor to play a Netflix stream; it has a dedicated block of silicon for that, and it's why your battery doesn't die in twenty minutes.5:10 Now, here's where the analogy holds.5:12 In all those cases, you trade flexibility for a massive leap in performance and efficiency on a single, well-defined problem.5:20 AMD is betting that some AI models, like Llama 3.1, are becoming stable and ubiquitous enough to be treated like a video codec.5:29 They're betting the model is the new standard.5:32 But here's where the analogy could break, and break badly.5:35 The SHA-256 algorithm for Bitcoin mining doesn't change.5:39 The H.264 video standard was fixed for years.5:42 AI models?5:42 They evolve in months, not decades.5:45 An incredibly fast, efficient chip that only runs Llama 3.1 is just a very expensive paperweight the day Llama 4 becomes the new standard.5:53 So AMD's bet isn't just on a chip architecture.5:56 It's a bet on the pace of AI progress itself slowing down, or at least stabilizing around a few key models.6:03 It’s a high-stakes gamble.6:05 Now, let's look at the other side of the coin.6:08 While the hardware is getting hyper-specialized, the process of creating software is getting...6:14 fuzzy.6:14 This brings us to that essay, "Taste Is All That's Left." The author's argument is simple and powerful.6:21 For decades, the barrier to creating software was the effort of creation itself.6:26 You had to know the language, the frameworks, the APIs.6:30 There was a floor.6:31 You couldn't just will a program into existence.6:34 That floor, the essay argues, is now gone.6:37 With tools like GPT-5.6 and Copilot, you can generate plausible, "good enough" code for almost anything.6:43 The cost of producing a function, a class, or a whole script is approaching zero.6:48 So if creating is free, where does the value go?6:52 It moves to selection.6:53 To judgment.6:54 The essay calls it "taste." That's the word for the verdict you reach when an AI gives you three plausible versions of a function, and you know, somehow, that two of them are wrong.7:05 Maybe they're inefficient, maybe they miss an edge case, maybe they're just...7:10 ugly and hard to maintain.7:12 That instinct, that discernment, is now the most valuable skill.7:16 We've seen this pattern everywhere.7:18 When the printing press arrived, the value wasn't in being a scribe who could copy a book by hand anymore.7:25 The value moved to the editor who could pick a good book to print, and the publisher who could market it.7:32 When digital cameras put a camera in every pocket, the value wasn't in just taking a photo.7:38 It was in having an eye, a sense of composition, a taste for what makes a compelling image.7:43 The analogy holds almost perfectly.7:46 As AI floods the world with infinite "good enough" content—whether it's code, or prose, or images—the premium on human curation skyrockets.7:55 The bottleneck is no longer production; it's perception.7:58 The danger, as the essay points out, is that "good enough" can dissolve the incentive to get better.8:05 Why spend a week crafting the perfect algorithm when the AI can spit out a decent one in five seconds?8:11 So what does it all add up to?8:13 You've got hardware becoming incredibly specific and fast for certain models, and you've got software tools that can generate infinite variations of code.8:23 You, the developer, are sitting right in the middle of this massive polarity shift.8:28 Your job is changing.8:30 It's less about the tactical, line-by-line act of writing code, and more about the strategic act of making choices.8:37 Which model do you bet on?8:39 Do you architect your system for a general-purpose GPU or for a hyper-specific chip like Taalas's?8:45 When your AI assistant gives you three options, which one do you pick?8:49 Why?8:50 Can you defend that choice?8:51 That's taste.8:52 It's what the founder of Herdr did.8:54 He tried the existing tools, his taste told him they were wrong for his workflow, so he built one that was right.9:02 That act of judgment was the seed for a project that now has twenty-five thousand GitHub stars.9:08 Contrast that with the Google Pixel bug.9:10 That's a failure of taste at the product level—an inability to imagine the real-world context where a feature would be more annoying than helpful.9:20 This week sets up a new reality for every single person building technology.9:25 The tools are getting unimaginably powerful, but they're also getting more opinionated, more specific, or more dangerously generic.9:33 Navigating that landscape requires more than just technical skill.9:37 It requires judgment.9:39 The most valuable line of code you'll write this year might be the one you tell an AI not to.
