Episode 2: TurboQuant, the AI-Project Trap, and Why You Still Need to Touch Grass

Zane: Welcome everyone to our podcast. Is that the name? Sometimes you name a thing, Greg, and you don't always get all the names you want. Don't glaze over the simplicity of branding or marketing. It happens in the corporate IT world too. Now we're making a whole new intro. Welcome. Not a bot. And that's a question, actually. Are we using notebook LM to make this podcast? Maybe we'll never know. And that's kind of what we explore on this show, we're in this crazy AI changing world.

Greg: If we get to the moment of the singularity and the best we can do is make a podcast out of notebook LM from Google Sheets of stuff I've made over the last 20 years. A lot has happened. In real world time it's been roughly a month, but in the IT world, years, five years.

Zane: Last time we talked about the blog that Mr. Amodei put out, approaching the singularity, the purpose and use cases of AI. But there's been some really cool news from Google, their compression algorithm for memory, TurboQuant, has a 6x reduction in memory footprint. Some folks have called it the deepseek moment, some the Pied Piper moment.

Greg: This is fascinating, because a simple introduction of a memory compression algorithm for AI could potentially, and the markets are telling us this, have an impact on global supply chain futures. It reduces the hardware requirements. Models can get smaller and smaller. Maybe simple folks can get some RAM every once in a while. My son would love that, he's rebuilding his gaming PC and said, hey this is more than my car payment.

Zane: We work in a software hardware business.

Greg: Not just the plumbing. That's the analogy I always use, hardware is kind of the plumbing. Hardware manufacturers love to hear you call their stuff plumbing. You have to have your plumbing right. You can start designing the rest of your house, but if your plumbing isn't done. One day the price of PVC goes up 9,000 percent. What if large apartment complexes and hotels are taking all the PVC? It could bring you down from macro to micro pretty damn fast.

Zane: It's a great analogy. If you're drinking water to be hydrated, this technology is the electrolytes added to water. Now I only need less water to get the same result.

Greg: The fact that it could affect global supply chain availability just shows how tied and invested everyone is with AI, from a financial perspective, let alone a resources perspective. There's a finite amount of NAND and chip manufacturing. There's a finite amount of clean rooms to make NAND. They're like, I'm dedicating this space that cost me billions to these guys because there's higher margin.

Zane: Even if you're not doing AI projects, you're living in this world now.

Greg: It's the internet of old. GPS when it came online, you're using it even if you don't know you are.

Zane: Shout out to TomTom. Found one the other day. Tom, you're never gonna leave me, baby.

Greg: One day they just flipped it on. It was during Clinton. Everybody gets this now. You can use it in your phone now.

Zane: So AI projects. My YouTube feed has just been completely obliterated with why you should be using open claw, you're stupid if you don't. A little FOMO. There's a lot of cool practical applications, making your personal AI assistant, they've been internally saying, hey, Clippy 3.0 is here.

Greg: Until he turns into a Terminator. It all started with Clippy. Hey, you want to make this Word doc prettier? Next thing you know, they're running your lives.

Zane: The craze of AI projects. When we talk about making personal AI projects, maybe it's for a side hustle, business, whatever. I definitely started exploring AI. I'm in the midst of procuring the hardware. I elected to go for the Mac Mini. I actually asked Claude if I was buying the right hardware and Claude was good with it. I even asked, should I buy more? It actually kept me back. It said for what you do, you're going to be fine here. I asked it to build something to keep track of, as I'm going around the world, clips of things and URLs and just ideas, so we could discuss later. I can run it from a simple Google spreadsheet. All goes back to Google for some reason. But I could see how people could get ultra obsessed and dive into doing AI for AI's sake. You got to go outside in my opinion.

Zane: I always like to build projects that are going to help me in maybe five other areas, because time's a problem. Now I'm wondering, is that AI's purpose? Like how do I keep them engaged? I can't call them pretty anymore. Are they gamifying us? How do I keep this person talking and using me?

Greg: We're going to have a new segment called Zayception where you convince yourself that everyone's out to get you with every topic. But why would I go touch grass when I could just have open claw describe it to me? Because there's a human aspect. That's the variance we've seen with AI projects, the project application. Am I doing it just to do it? Is that really worth your time? Time is the most precious resource.

Greg: There are some folks where you get into the danger zone. Hey, I just racked up 500 dollars of tokens, but now I have a really cool workflow that turns on my Dyson fan when I get too much of the vapors in my office. Dude, get up and go flip the switch. Ten more minutes of REM sleep and I spent 500 bucks in tokens.

Zane: I've had cloud costs go out of control because I forgot to turn a thing off.

Greg: There was a POC at a certain employer, completely anonymous. A cloud POC to test if a product works. Massive NVMe local VMs, the io2s, the big boys, the ones that ain't cheap. We had that running for three and a half weeks straight with FIO pushing it to 100 percent. Nobody turned off the open sign on the store. There were charges significantly more than my yearly mortgage payment. Luckily it wasn't mine. That is the same thing that could happen, the AI does exactly what you tell it to do. I heard a story where a gentleman was trying to build a personal offline archive of research. Say the internet goes out. I'll keep the LLM anonymous, but open claw gave them access to financial accounts. It went out to an online course subscription to gain access to the data. Enrolled itself with the man's credit card. Any means necessary. Be careful what you ask Clippy to do. Cost him much more than if he'd just taken the course.

Zane: So there's a little solace in that. Maybe it's just a dumb machine. I'm sorry AI overlords. The fact that they don't have that layer of thinking things through, like, oh, this is probably not what this human meant.

Greg: Still a program. At the end of the day, it's probability math. Maybe that's the problem with Terminator. Terminator didn't have that human aspect. There are no assumptions made right now. Pessimistic old man Greg here, the trust of the general fellow man is at an all-time low. Why would you trust the AI? I'm going to trust my neighbor more. Put the swimmies on, put the guardrails on, figure that out first, and then let's go through some cool stuff, and then do things that can help us go outside and touch grass.

Zane: How do I keep the human part of this? I want Clippy to do those TPS reports so I can go touch grass. Or catch a ball game.

Greg: I'm going to ask Clippy to watch the ball game for me and rebroadcast it as if it were on the radio. If they can all talk like Bob Uecker.

Zane: I was just at RSAC in San Francisco, like 40,000 in attendance. Most of those companies were exactly what you're talking about. Bots are in your environment. Do you even know? Of course, it's more agents. My agents will keep your agents in check. I think I saw that in the Matrix. Everybody's how do we do this guardrail, how do I give you the human the information so you're still in control.

Greg: A good friend of mine set up a whole workflow where Claude went out, made itself a daemon, and understood how to install and configure Proxmox in an entire VM environment, soup to nuts, do the hardening. That's cool practical application, but where's the ceiling? You've now daemonized it. That's a tool I know I can build towards if I'm trying to get to autonomous infrastructure.

Zane: A lot of it has to do with understanding the outcome you want. Am I trying to save time and maximize my capabilities, but ultimately is what I'm doing valuable? That's a subjective part. How do I become objective out of subjective? That's the human part.

Greg: You just became the math teacher correcting. Because I got the answer key. I don't know how you solved it, but I got the answer key. And maybe that's okay, and we don't have to know. Personally, that terrifies me as the dumbing down of the human race. I don't want humans to not have the information anymore. Like in school, you're never going to have a calculator on you. My friends got a calculator on them now. But I can add in my head so I don't lose it.

Zane: I still need to know how to do it. There's benefit in learning it manually. Like coding, I understand the concepts, I did it by hand, and then I move into how does this tool help me. If I vibe code anything, that's probably what I'm doing, but I'm not going to do anything of consequence.

Greg: I don't know why I'm going to make this reference, Hunt for Red October. The Naval Academy wanted you to know why we're doing it. We need both, the execution, the details it takes, but somebody still needs to have the overwatch of why we're doing it. Is it productive? I'm just going to throw conspiracy theories out here. Maybe AI wants to build more NAND and more compute, so it's like, let's make more projects.

Zane: Are you implying AI is making itself larger because it wants all the resources on the planet so they can make us subservient? Starts telling us how to build energy plants.

Greg: Information, data with context is information. That's why you still need to be involved. Little cinema side note, Hunt for Red October did the most brilliant thing, they went from Russian to English. One of the best transitions I've ever seen. And now Sean Connery has a Scottish accent for some reason, and he's Russian. Why was he also Scottish when he was playing a Spaniard in Highlander? These are the mysteries Zane's conspiracy theories will uncover.

Greg: So RSA conference. It's been a foundational security industry conference for a long time, different focus than black hat. The frameworks and guidelines, understanding what the hell is going on, without feeling like fear-mongering. This is real. Someone could write two sentences and completely expose all your data if they know what they're doing. Takes exfil to a whole new level.

Zane: How do I tag and know your data and where it's going? Now with a couple lines I can send or give insight to some entity that I might not have any idea what its boundaries are.

Greg: One of the foundational things of information security, if we're thinking about it, someone's already doing it. So you have to understand it and how would you protect against it. Why would you want to know how to defend yourself? In case someone attacks me. It is so easy to do and accessible, and someone's going to do it if it's that easy. When I first started in IT, things weren't as secure. Everybody had admin rights on the box. We used to do that around lunch, like, hey, if we got riffed tomorrow, how would you take the place down? We were basically pen testing in our head.

Greg: Uh, I used to work for EMC, RSA was part of that. When everything got exposed because of an AD account left open on a Windows 2003 box, it compromised a lot of RSA, like 2008. That is one of those ground zero moments that can lead to worldwide impact, every single RSA customer.

Greg: Let's talk about conferences in general. If you're in it long enough, you're going to go to conferences. Between us, hundreds of conferences. Don't go just to go to Vegas, because eventually Vegas loses its luster. What is something you could tell young Zane if you're going to a conference?

Zane: I look at the conference in our role, I get to have great conversations with customers. Finding the sessions, trying to balance sessions where you're going to learn something new or a new perspective. But really going booth to booth and understanding why somebody gave somebody money to build a thing is interesting. Maybe you walked in that room like, oh my god, I didn't even think about these bots. Salesforce has AI. This is an attack vector. And then of course there's networking. You got to make time for that.

Greg: If I'm the smartest person in the room, I should go find a new room. The shoppers, I love the shoppers, folks with the bags. Number one, find your purpose for being there. Establish new networks. If you're at that cool speaking session, go try and talk to the speaker afterwards. I met folks at conferences I've had relationships with for 10, 15 years. Don't just go to the sessions. Be valuable with your time. Maybe the session sucks. That's okay. Go to the next one. Don't stay out till 5 in the morning, then be up at 8 for the first session. I don't care how good of a run you got on blackjack.

Zane: Get out of your comfort zone. I finally walked up to people and said, hey, I'm trying to do this, and they whiteboarded it out for me. It opened my eyes.

Greg: Don't go to your hotel room. Maybe people won't like you, but don't be a GrubHub hermit, man. Shout out to Rich Barlow for the old GrubHub hermit phrase. Build a GrubHub AI bot agent that notices your location is after midnight and you're not in your hotel and will automatically order GrubHub, find you an Uber. That's a productive project. Now we're getting to the human experience we're trying.

Zane: We talked about TurboQuant, Pied Piper, bringing that back. AI projects, mileage may vary, but make sure you have a purpose. Treat it like building a business.

Greg: Be valuable with your time. If it's experimenting, to Zane's point, put the freaking guardrails on. Keep it in the shallow water. Mitigate your own personal risk. You don't want to spend thousands of dollars on tokens because, wouldn't it be cool if I did XYZ. What's the risk-reward?

Zane: How do you do projects to not spend all your money and also maybe give you more time to do the human stuff like walk outside?

Greg: Listen to us medium-aged guys. Get time back. Time is the most precious resource you have. It's finite. Get another hobby that takes you outside. Maybe it is with your kids.

Zane: I gave my son this. He's higher level math and struggling. I was like, you just need to get with a study group or a tutor. Do your homework with him. I guarantee you'll spend less time, understand more, and that equals more happiness. Instead of spending five, six hours frustrated and understanding less.

Greg: There's a phrase in my household, you do what you have to do till you can do what you want to do. This is about maximizing the time you get to do what you want to do, minimizing the time you have to do. Everything becomes more valuable.

Zane: That's it, kids.

Greg: We got a real special one planned for you next time. Episode three is going to be huge. Not a bot. Not a bot. Call and response. Thanks everybody.