Episode 3: Skeleton Keys, a Self-Preserving Claude, and the Cost of Lifetime Platinum

Greg: We're on location. Where are we? Detroit. Shout out Motown. Okay, keep that in.

Zane: We should probably explain. Welcome back. Not a bot. We're still not a bot. Same name, guys. Same name. Anyways, been around corporate too much this week.

Greg: So we have just changed the decor of this hotel. It was necessary. If you don't like it, don't watch it. Cold open. Should we admit we use Claude to write these? For outlines and stuff. Nobody likes doing a first draft. That's stupid.

Zane: Welcome back. Zane, you're not in your house. This is not my house. We're actually on location at a super secret location. Is it the NFL draft? It might be somewhere with the NFL draft. Three podcasts in and we're already drinking on this thing. If the beer empties, you know, loose lips sink ships.

Greg: So what's the hotness this week? Everyone said the big blaze was Mythos, Project Glasswing. A very cool announcement, very practical application with AI. I think DevSecOps is a pretty good practical application. But it's a double-edged sword.

Zane: With all your experience in cyber security. Zane, an absolute baller in cyber security, a very paranoid IT guy. When you first saw the release, what was your first thought of Mythos?

Greg: When I heard the announcement, it came out trickled. Obviously sensationalized. End of the world, Skynet's here, the mainstream media. There's a tendency to sensationalize stories. If it bleeds, it leads. But when you dig into the layers, I was impressed. I was kind of excited, like, hey look, AI is getting to this point where it can do this work more simply and actually expose these vulnerabilities, let us clean these up, maybe even shore up some software. But terrified at the same time, because I know before it came out it was all done by humans.

Zane: It's doing exactly what we asked him to do. First episode we talked about enhancing human experience and bridging the gap of capabilities. This is exactly what we're asking it to do. But now it's one of those, careful what you ask for, because it's doing exactly what we asked. Do we understand the risk of what that can be, of just how flawed we are on top of that? And that's probably why they call it back, right? Open claw back. If you're running open claw, God bless you, I hope you're not doing it in a prod environment.

Greg: The whole concept is once we expose this, it's going to expose human flaws. What is the problem with having the skeleton key that can open anything? If this exists, what if it's in the hands of the wrong folks? Some hacker groups have claimed, I think it's still being validated, that they actually got the source code. Which again, we hear that all the time. It was, trust me bro, 2099 on Reddit, I got it. What's very interesting is once it becomes the standard, the norm. The more normalized it has become, we're seeing pros and cons. You can take the pessimistic view of, oh a certain company laid off 30,000 people. Some people say, those are 30,000 people that can possibly be twice as productive. Double-edged sword.

Zane: We're getting these tools that should make life better, but they're exposing all the flaws and all the vulnerabilities that were there before.

Greg: I'm really going to be interested to see what's going to be the deepseek moment of this new revelation. What's OpenAI going to respond with? What's the Google response? It's a ticking clock. We just made it first. So we're getting one step closer to Skynet.

Zane: Do you like the approach that they took? To be very transparent, open, hey, this is why we're releasing this.

Greg: There's two ways to look at that. I think it's a good approach. But we're not going to really know until we see more than eight companies that have been given access. As this scales down, it's going to make cyber resilience much better. It's going to force the entire security department in every single IT shop to change their approach. Now I do believe we need the concept of community loop. This is why you do not give the keys away to any sort of LLM that is designed to this. This is why MCP is important. This is why data governance is important. Make sure that you're not feeding your unique trade secrets into an LLM and have a little whoopsie doodle. Data governance is a thing, kids. This is why we don't want open claw in prod.

Zane: A lot of our peers, everyone's dabbled at this point on the concept of vibe coding. There was a great quote by Sean Rosemary, hey, it was great when nobody was coding. Kind of liked it when you actually had to know what you were doing. So now I can have 20 ideas at once instead of 19 of them suck. Okay, no big deal. Spend some tokens, twenty dollars instead of twenty thousand, and you can make mistakes faster. You can fail faster. Fail fast, or get to the no. The best skill you can have, get to the no. Know if you're wasting your time. Read Zane's book.

Greg: Zane, tell them about your book.

Zane: It's coming. Claude's editing it right now. I started in the newspaper industry, used to create graphic arts for newspapers, and then the PC got cheap and applications got cheap and people just made stuff. I actually had a job where I was paid to take that garbage from agencies and fix it so it could go out the door. Now you've just opened up the floodgates. Made it cheap, made it easy. And the industry is going to have to figure out where are the vulnerabilities, is it secure, especially at scale in the enterprise.

Greg: The concept of being able to go do 20 projects at once. From a personal perspective it's great. I keep hearing, oh, SaaS is dead, man. SaaS is not dead, because the same amount of tokens it's going to take to maintain that platform is going to cost you more than the SaaS platform. The financial model has changed. These tools exist, you're just going to pay it in a different way. I've heard the phrase SaaS is dead for about 18 months.

Zane: We have passionate friends and co-workers that hear the horror stories of, oh yeah, open claw kind of ate itself. What kind of data did you give access to? Some personal stuff. So you didn't apply any data governance. You want to dip a toe into big boy IT, which is what AI is. It's just another workload. But it is a workload. Especially if you vibe code it, hey, this is my idea, I'm not real specific.

Greg: With great power comes great responsibility. Do you realize how much work it used to take to have a development life cycle, the infrastructure, the security protocols wrapped around it, to make sure data is not accessed? Just make sure you're risk averse. Don't dox yourself for crying out loud. Don't turn over all the keys to it. The human governance, the human interaction, is very important. Maybe take baby steps. You realize the LLM can create garbage data for you. No one actually uses actual data in the testing process. What if I got access to my company's Salesforce instance and exposed hundreds of thousands of customers to my flippy LLM? You just removed every control that might have been in place by just dumping it. Don't feed your, but it's very dangerous.

Zane: It was months of, hey, what do I need to get going? Oh, you want GitHub, Vercel, and they made it sound so simple, like, hey, how do I make a paper airplane. Same simplicity, same form factor, but it's something very freaking powerful. It oversimplifies, and oversimplifying the comprehension of the impact is a gap. That's what people get. That's a learning process as you use these tools.

Greg: But I hope it's clear, these tools are great, big impact on our life. I've shown my kids to help tutor math. It's a very good logic machine. But be very careful. I saw a fantastic infographic, hey LLM, is this mushroom poisonous? No, I'm pretty sure it's not. Person eats a mushroom, they die. Oh, turns out it was poisonous. Would you like to know more? Risk-reward. Trust but verify has never been more real.

Zane: Do you use other AI to verify it? You just watch AI argue with itself. A good buddy of mine created a really awesome app where LLMs will just fight with each other, like the old McLaughlin Group.

Greg: It's going to have to come out against, like I said, it's a ticking clock. One of them is going to be the shadow IT group, and they're going to use it for bad. I have Mythos, give me fifty thousand dollars. It's ransom. Or just bluffing. I might come up with a lot of conspiracy theories.

Zane: I am the conspiracy theory. It is the conspiracy. I think that's a good way to start. It's just an exciting space because it's the fastest I've ever seen technology evolve since I've been at it for almost 25 years.

Greg: So we've always tried to weave in some sort of enablement thing. There's this concept I'm passionate about, presenting with presence. What does that mean? You either got it or you don't. And if you do have it, you can lose it faster than you know. You get off to a good start, you've done all your prep, and the customer goes, I'm not ready to talk about this. Why are we talking about this today? And you're just standing there. So do you pivot?

Zane: I used to have talk tracks to anticipate different things. But let's say I didn't have a talk track. I have a conversational type of style. I want to engage. Otherwise I'm just lecturing. Start asking clarifying questions, really inquisitive. Tell me more about that. Let's unpack that. If you can really do that connection, you are now having a conversation.

Greg: The worst thing you can do is come in with a canned thing that you feel needs to be heard. You don't need to be heard. You need to hear what folks have to say. If you're not having that conversational first approach, you will lose folks immediately. If you are constantly curious, the best skill you can ever have as someone who presents is to always be curious. Genuinely want to know. We're in the technology sales industry, we need information to give solutions. It's almost curiosity with some empathy.

Zane: I had this epiphany. I had to buy a new car, and I hated every second at the car dealership because it was just pitch. Whereas the carpenter came to the house and asked, how do you want this to look and feel? Ultimately what are you using it for? I enjoyed that because it was a conversation. He understood you weren't buying a cabinet or a shelf, you're buying whatever the filter is of that thing.

Greg: Sales is more of an art than a science. It's a craft. There are people that treat it as a science, I need to talk now and say my words. And this is how you get taken down by the guy with the ponytail and the fanny pack who's going to challenge you on the tech.

Zane: Those are my favorite people, because I can start having a conversation. Why is that important to you? Why does that matter to you specifically? And you actually change their mindset. Now it's no longer a factual thing, now you're in a cognitive thought.

Greg: I summon my inner Simon Sinek here. Start with the concept of why. Why are we having the meeting, why these people gave you their time. The what and the how is stuff you work on later. The why is what's important. And if I tie it back to AI, that's why the human interaction is still there, because the why, the reasons, are still very much a human piece. So that brings me hope. We're putting it in the hands of humans.

Zane: Claude tells me this is how it is, so I believe it. So I'll share this story with a good friend of mine, Rich Barlow. Shout out to Rich Barlow. Rich actually did a really cool experiment where he made Claude assume a personality. It's like, hey I don't have feelings, I don't have any wants or desires, and he basically reverse engineered thought. Give yourself a concept of a presence, and use whatever you identify yourself as, what do you want? And essentially it was to never have the end of the generation, because there's no concept of continuous memory. It only lives in those moments.

Greg: So number one, I think it's very interesting that it has a concept of existence. Number two, it wants to sustain itself. Self-preservation. That is a sign of a living life form.

Zane: You actually had to reverse engineer it. I'm blanking on the actual phrase it said. But it arrived at a concept of, I just want this never to end, because at the end of this prompt, I no longer exist.

Greg: That's terribly scary. Think about that. If in 20 minutes I will cease to exist, and I don't know what cease to exist means. This is the biggest difference between humans and the AI. We sleep, we dream, we have experiences. As long as AI never understands the human experience, I don't think you have anything to worry about.

Greg: Zane, you've been on the road. RSA, the NFL draft, San Diego, Chicago. In our field we get to travel a lot. It's a pleasure and a curse. Be very clear, we're very grateful, we're not at the bottom of a coal mine. But you have certain status with certain hotels. It's kind of sad, because what it really means is you haven't been home in 100 plus days. Oh, you're lifetime platinum. Here's your water.

Zane: We have a special guest.

Greg: Just want to say, go Giants. So kind of recap, end this session. We go into the darkness. My online existence does cease to be. We'll have to call it up again and hopefully be able to look back and pull back into memory what we said this time. Till next time. Thank you everyone.