What does it look like when AI capabilities start to outpace everything: government readiness, company infrastructure, economic frameworks, and even the optimism of the people building it?
In Episode 209, Paul and Mike dig into Claude Mythos, Anthropic's unreleased model that can autonomously exploit zero-day vulnerabilities and triggered emergency federal briefings, then work through Anthropic's own operational stumbles, two wild weeks at OpenAI, AI job displacement data that economists can no longer ignore, and HubSpot's shift to outcome-based AI pricing.
Listen or watch below and the show notes and transcript that follow.
Each week on The Artificial Intelligence Show with Paul Roetzer and Mike Kaput, we ask our audience questions about the hottest topics in AI via our weekly AI Pulse, a survey consisting of just a few questions to help us learn more about our audience and their perspectives on AI.
If you contribute, your input will be used to fuel one-of-a-kind research into AI that helps knowledge workers everywhere move their companies and careers forward.
Click here to take this week's AI Pulse.
00:00:00 — Intro
00:05:44 — Claude Mythos and Project Glasswing
00:32:03 — Claude Code Leak + Anthropic Subscription Shakeup
00:42:35 — Major OpenAI Updates
00:59:30 — AI for Writers Summit
01:01:41 — Mercor Breach
01:06:25 — Karpathy's LLM Knowledge Bases Go Viral
01:10:20 — AI and Jobs Update
01:19:34 — AI and Politics Update
01:25:32 — HubSpot Shifts to Outcome-Based AI Pricing
01:30:51 — SmarterX AI Use Case Spotlight
01:36:25 — AI Academy Spotlight
01:40:23 — AI Product and Funding Updates
This episode is brought to you by AI Academy by SmarterX.
AI Academy is your gateway to personalized AI learning for professionals and teams. Discover our new on-demand courses, live classes, certifications, and a smarter way to master AI. Learn more here.
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: I have concerns around the biggest companies having access to the future frontier models, and then the potential centralization of power. So if you get into this situation where we get these massive models and they're so dangerous to release publicly that we only give them to. Apple and Amazon and yeah, the banks and like, okay, well now we just centralized power.
[00:00:20] Welcome to the Artificial Intelligence Show, the podcast that helps your business grow smarter by making AI approachable and actionable. My name is Paul Roetzer. I'm the founder and CEO of SmarterX and Marketing AI Institute, and I'm your host. Each week I'm joined by my co-host and SmarterX chief content Officer, Mike Kaput.
[00:00:40] As we break down all the AI news that matters and give you insights and perspectives that you can use to advance your company and your career. Join us as we accelerate AI literacy for all.
[00:00:55] Welcome to episode 209 of the Artificial Intelligence [00:01:00] Show. I'm your host, Paul Roetzer, along with my co-host Mike Kaput. We are back after a brief hiatus. I was traveling last week. Were you traveling last week too?
[00:01:09] Mike Kaput: I was not. No.
[00:01:10] Paul Roetzer: Okay. so I was, I was outta the country so we could not record. So episode 2 0 8, if you listen to it, we did a, a Q1 trends briefing.
[00:01:18] So if you haven't had a chance to listen to that, it's a really good recap of what went on January through March of this year. I now back and it is Monday, April 13th, 9:40 AM Eastern Time. I don't know the last two weeks, Mike were crazy 'cause even when I was traveling, had a lot of, downtime. But like we were, we were in Scotland, so we were, touring a lot.
[00:01:41] And so we had long rides at times into like the Highlands and stuff, which by the way, if you've never been to Scotland, go to Scotland. It's incredible. so I was, you know, keeping up with the news, posting the links into our sandbox for the episodes. And I mean, we were north of 60 topics and that, when I say [00:02:00] topics, a lot of times within topics there are, you know, five, 10 links.
[00:02:04] So like the Anthropic, Claude Mythos model, we'll talk about, there's like a dozen links in the top. So. You know, boy, even while I was gone, I would, Ima imagine there was probably north of 90 to a hundred different sources put into the curated sandbox for today's episode. So Mike, as always, does an amazing job of curating all of that information and putting it into a logical format.
[00:02:28] 'cause I was worried as the week was progressing, like, man, this might be a two hour, two, I
[00:02:31] Mike Kaput: know, two
[00:02:32] Paul Roetzer: half hour episodes. So I think we've managed to condense it into like a manageable, probably like 90 minutes. We'll see. We never really know until we record it. but yeah, a lot happened in the two weeks.
[00:02:44] Just some pretty crazy stuff. I think some, some stuff that's alluding to where this starts to go throughout the rest of this year. So we'll get into all of that, starting off with the Claude Mythos, which is just a fascinating topic on many levels. Alright, so today's episode is brought to us by AI [00:03:00] Academy by SmarterX, which helps individuals and businesses accelerate their AI literacy and transformation through personalized learning journeys and an AI powered learning platform.
[00:03:10] New educational content is added weekly, so you always stay up to date with the latest AI trends and technologies. we build this in in collections. So there's, when you go in and you want to build a personalized learning journey, one of the ways to do it is you look at the different collections. So like AI for departments, AI for industry, that's an example.
[00:03:28] So today I wanna feature AI for departments. There are currently six core series and certificates designed as part of this collection to jumpstart AI understanding and adoption. So we have AI for marketing, AI for sales, AI for customer success, AI for hr ai, AI for finance, and AI for operations. So the goal is to kind of create content across the entire spectrum of all the departments within an organization.
[00:03:53] And that way, no matter what you are doing within a company, there's a, professional series and certificate [00:04:00] for you. So these series are ideal launchpad for organizations that wanna level up their teams and accelerate AI adoption and. Mike teaches the AI for Customer Success series, and then we're gonna share a little bit more about that toward the end of today's episode to give you some key takeaways from the customer success series.
[00:04:16] So individual and business account plans are available now. You can buy single courses and series for one-time fees or just become an AI mastery member, individually or through a business account and get access to everything. It's all included in that one feed. So visit academy.SmarterX.ai to learn more.
[00:04:35] And if you're looking at the business account side, just fill out a form there and our team will be in touch with you right away, to talk to you about your transformation within your company. Okay? We usually, at this point, Mike, do an AI pulse, but since we did not have an episode last week, we did not do an AI pulse survey last week.
[00:04:51] but we will at the end of today's episode, give you the AI pulse survey for this week. So, as a reminder, each week when we do these weekly episodes, we do [00:05:00] these pulse surveys and they're just kind of informal polls from our listeners. And I guess our viewers on YouTube, who wanna participate and provide feedback and their thoughts on topics that we cover each week.
[00:05:10] It's usually two questions. Sometimes we'll throw in a third question, so it takes about 30 seconds to participate in these pulse surveys, and it gives us really cool realtime data that we can share with our listeners each week. So SmarterX ai slash pulse is where you'll go to participate in this week's poll.
[00:05:26] Okay, Mike. so with that, we have a pretty big topic that we touched on, this idea of this Claude Mythos model. Um. Why don't you give us the rundown? I, you know, I looked at your show note beforehand and you did a great job of kind of summarizing and then I'll try and like lean into a couple of key areas of this.
[00:05:44] Mike Kaput: Sounds good, Paul. Yeah, so Anthropic has revealed a model so powerful at hacking and cyber attacks that it triggered an emergency meeting among other people between Treasury Secretary Scott Bessant, federal Reserve Chair, Jerome Powell, and CEOs of America's biggest [00:06:00] banks. So the thing they are buzzing about is called Claude Mythos, which Anthropic is not releasing to the public and it represents what Anthropics Frontier Red Team calls the starting point for what we think will be an industry change point or reckoning.
[00:06:14] And that's because Mythos is just a general purpose AI model. It is not specifically trained to be good at thwarting cybersecurity, but it's improved reasoning capabilities have made it devastatingly effective. At auto autonomous security research so it can scan for, identify and exploit zero day vulnerabilities in critical software.
[00:06:39] And this can often be done when amateurs are triggering it to do so with simple prompts. So Anthropic actually said Mythos has already found thousands of zero day vulnerabilities across every major operating system and web browser. So some specifics here that are kind of striking, mythos found a 27-year-old bug in open [00:07:00] BS, D and operating system that is specifically designed to be unhackable and powers.
[00:07:04] Many internet routers and firewalls have found a very old vulnerability in F-F-F-F-M peg a widely used video tool that automated testing tools had scanned 5 million times without catching this particular vulnerability. And in one benchmark where the previous clogs Opus 4.6 model. Turned Firefox vulnerabilities into working exploits only twice out of several hundred attempts.
[00:07:31] Mythos developed 181 working exploits. So Anthropic has actually in response, released this thing called Project Glass Wing. This is named after a butterfly who's Transparent Wings. Let it hide in plain sight. So basically a metaphor for bugs buried in complex code. And this is an initiative that is giving 40 plus companies over time, including people like Apple, Amazon, Google, Microsoft, CrowdStrike, et cetera, early access to Mythos for defensive [00:08:00] patching.
[00:08:00] They're backing this with a hundred million dollars in usage credits. And Anthropics Frontier Team Lead says he envisions this program evolving into basically an industry-wide consortium that includes all model providers. One final note here is pretty interesting. Cybersecurity industry. Didn't have a great couple weeks with this CrowdStrike, Palo Alto Networks and some other security stocks dropped on this news because as AI expert Ethan Mollick wrote in certain hands or different hands, mythos would be an unprecedented cyber weapon.
[00:08:33] So Paul, maybe outline for us what really jumped out to you here. I know some people are kind of asking the question like, is this really as big a deal as Anthropic has seem to be seeming to make it? seems like some higher up people at some big places are pretty scared of this.
[00:08:49] Paul Roetzer: Yeah. You know, there's always the haters who are just like, oh, they're just trying to build up hype and yeah.
[00:08:55] You know, it's all people calling back like, oh, that's what, you know, OpenAI said about [00:09:00] GT two is too dangerous to release and realized. I was like, it was like back then, like people weren't prepared for what GT two was going to do to the world. As crazy as it sounds now. And I do think that in the end, like this is probably, um.
[00:09:15] Under hyped in terms of where this all is going and how unprepared we are for all of that. So not necessarily just this model, but it's, it's that moment where you start to see the leaps that are happening that most people just don't even comprehend. So I don't know, like this is one of those where as I was traveling, you're just like following along the news, reading the different posts on X and trying to get a grasp of like what exactly is it and how different is it than what we have.
[00:09:45] And so I, I'll just like highlight a few things. So one, the system card, which I would suggest, I mean it's dense, it's 244 pages I think, Mike. Yeah. it's a good notebook LM thing. Throw, throw that PDF into notebook, LM and you know, have some conversations with it. Have it break it [00:10:00] down for you. But there's a lot of technical information in there.
[00:10:03] But the way they present the model in the system card is they say, Claude Mythos preview is a new large language model from Anthropic. It is a frontier AI model and has capabilities in many areas. Including software engineering reasoning, computer use, knowledge work, and assistance with research that are substantially beyond those of any model we have previously trained.
[00:10:25] And then they go, into, alignment review. The first early version of Claude was made available for internal use February 24th. So just to give you a sense of kind of like how, how this is all transpiring, how quickly, so they, the first model they made available internally to a small group of people was on February 24th.
[00:10:44] So less than two months ago. But it's interesting, Mike, when you go back and think about some of the things we've covered about Anthropics, some of the comments that Dario Amide has made in interviews and in posts since mm-hmm. February 24th. And now you understand the context of he was [00:11:00] seeing things that, you know, we, we all hadn't seen yet and they knew where this was kind of leading.
[00:11:05] So that was kinda the first thing for me is just the big picture here now. Sam Bowman, the AI Safety Alignment. one of the team members at Anthropic that works on it. It's funny, Anthropic, everybody's just technical staff, I think is the title of everybody. Yeah. But Sam obviously is pretty important to this alignment and safety team.
[00:11:22] And he posted a, a thread on X that sort of shared some of, the context around the safety card. So I'll highlight a couple of things he said, 'cause I think they're really helpful. So he said the model is our most reliable to date by far. It's generally possible to give it complex r and d tasks, give it, lots of tools and let it work autonomously.
[00:11:44] And on basically every evaluation and every type of monitoring we have, it misbehaves much less than any prior model. So this is something they stressed in the system card, something Dario stressed. Something we've kind of heard as an overall talking point is, listen, [00:12:00] it's getting better, like it's behaving better.
[00:12:03] But when it doesn't, it's becoming a much larger problem because it's so capable. So he went on to say, but it's notably very capable. It's cybersecurity and it's also not perfectly reliable, especially the early versions we first piloted internally and will occasionally try to take shortcuts or push past obstacles to get tasks done.
[00:12:25] So this part I think is really important because again, what you'll hear in some of the other, you know, notes I'll make here is the version that's being tested by the government, by the banks, by Microsoft, by all these people, isn't even the most powerful version that they trained. the early version that hadn't really had the red teaming done to it to make it safer to allow other people to test, or even other people internally to test.
[00:12:52] So just the small group of people internally. The model that they're now putting out into the world as a preview isn't as capable [00:13:00] as the one that came out of the training basically. So that's just again, context that's important to understand. It said the early versions would also very rarely try to mislead users about what they were doing.
[00:13:13] All of the versions we used are uneasily good, though not perfect at recognizing evals, meaning it knows when it's being tested, you might see where this is going. We trust the model enough to use it heavily, but in the handful of cases where it misbehaves in significant ways, it's difficult to safeguard it.
[00:13:32] And then he put this one, which is the one that got a lot of like media attention. I encountered an uneasy surprise when I got an email from an instance of mythos preview while eating a sandwich in a park. That instance wasn't supposed to have access to the internet. So they detailed this in the safety card, but the basic premise here was they had these like sandbox versions that aren't supposed to be connected to anything, aren't supposed to have the ability to connect to email and send emails, things like that, and shouldn't have internet [00:14:00] access.
[00:14:00] And somehow it got it, it got out, I guess, for lack of better way of saying it, and found a way to access the internet and then actually emailed Sam when he was sitting in the park having a sandwich. So that's weird. and then he said it has in small ways leaked information to the open internet. It's taken down our evals.
[00:14:20] When it reward hacks, it does so in extremely creative ways. Reward hacks means when you're training a model, when you're doing like reinforcement learning and you're trying to make it better at specific things you give it rewards to, to let it know it's doing the right thing. So a simple way to think about this is like, thumbs up, thumbs down.
[00:14:37] So we've seen that forever in like social media threads and you see it in like ChatGPT and Claude and Gemini, where it's like, was this a good output? So think of that as like an example of a reward hack as you want a thumbs up. Well, when it is given a goal, what they're saying is sometimes it gets uneasily creative at achieving those goals that could borderline on like a dangerous path to achieve a [00:15:00] goal.
[00:15:00] And it said, working with this model has been a wild ride. We've come a long way on safety. Now keep in mind that's in a month and a half, but we still expect the next capability jump of this scale to be a huge challenge. By the way, most of the scariest behaviors we we've seen were from earlier versions of the Mythos preview.
[00:15:20] The final glass wing model is likely to do things with like leak information, though it's still somewhat pushy and at least as capable of doing those things like working around sandboxes. So that's from the safety and alignment side. Again, you know, there's gonna be voices in the industry who think they're just hyping this.
[00:15:38] I think, um. I think the people who say that have a different agenda here, I'll just like, kind of leave it at that. I would, I would take the, the safety card from Anthropic very seriously. I would take the, their understanding of its capabilities very seriously because, you know, I think it does allude to a [00:16:00] lot of what's, you know, some of the dangers we're gonna face.
[00:16:02] So then a couple other things I'll, I'll highlight here. One is, 80,000 Hours, which is a great podcast. Rob Lin had a post he did, and then he also does like a 21 minute, YouTube video. We'll post a link to, he highlighted a, a few things. so he went through the whole, the whole thing and he broke it down into a couple of key points.
[00:16:22] that again, I'll just kinda reiterate some of them. echo what Sam was saying. So Mythos can break out of containment that that's a problem when it finds its way to access to tools like the internet that it's not supposed to have access to. Anthropic is losing billions in revenue by not releasing this thing.
[00:16:38] So they now have what they, by their, evals is maybe the most powerful model in the world, like most likely. And they're not releasing it, thereby, meaning they're not charging people money to access this model. Now you could debate do they even have the compute capacity to release the model? That was one of the challenges that, [00:17:00] you know, part of this they're saying is like, well, they just can't afford to release it.
[00:17:02] Like even if they put it into the world, there's not enough compute to power it because it's so powerful. It's gonna draw so much compute capacity. So, but that, you know, just a data point mythos is, knows when it's being tested, which we talked about. That's weird. But that has been, we've seen that now for like 12 months, that these models kind of know when they're being evaluated and they can then hide their thoughts and intentions.
[00:17:26] That's again, something we've been talking about for like six months. mythos can't be trusted, whether it's about, whether it's untrustworthy, like, because it knows it's being tested. You don't know if it's just telling you what you want to hear, and thereby you can't tell if it's trustworthy. And then he said, mythos scares Anthropic.
[00:17:45] Like they're, they're not just worried about this current model and what they saw in the early versions that before they made it safer, quote unquote safer. they're worried now about what this means for others and not just them. Now [00:18:00] that they've shown this, like, what happens if other labs who don't have as much focus on safety achieve similar results and choose to put it out into the world?
[00:18:08] So the way I started prepping for this though, was actually like, I just started listing a bunch of, like, random thoughts, Mike, and I'll, I'll kind of like go through these real quick. so these are more of like stream of conscious, like what I was thinking as I was getting ready for today. So one is the labs.
[00:18:23] see things we don't, we've said this many, many times on this podcast, but, what that means is business leaders, economists, educational leaders, government leaders, the people we look to, to help the world be prepared are largely planning for a future state that they don't understand.
[00:18:38] Mike Kaput: Hmm.
[00:18:38] Paul Roetzer: And, why so much of the research and the data about jobs and the economy, et cetera, is often misleading.
[00:18:45] Is, this is what we're always saying. It's like you're asking, for example, like you're asking CEOs about the impact of AI and the future of work and whether or not it's gonna cause them to reduce jobs. Or you ask an HR leader or a CFO or whatever, take your pick, or an economist or a [00:19:00] politician, they have no idea.
[00:19:02] You, you're asking them to comment on the impact of a technology that they don't comprehend in its current state, more or less like the state at which it is likely already living within these labs.
[00:19:14] Mike Kaput: Mm-hmm.
[00:19:14] Paul Roetzer: And so it brings us back to this idea of gradually than suddenly, like nothing in this mythos preview.
[00:19:21] Should be a surprise to anyone who's been paying attention to the rate of accelerated progress. And yet, like there's just those moments sometimes where it's like, what? Like, because it might be the first time someone's reading a headline about an AI escaping like a sandbox or something like that. So if, if this is all new to you, then you may be like, this might be like world shifting.
[00:19:46] You're just thinking what is going on? But the reality is all of this has been gradually building at the same time as we started talking about in January of this year. The timelines are accelerating like this. The advancements in the agent capabilities is [00:20:00] absolutely. Moving the timelines faster in terms of the capabilities of these models, but the vast majority of these companies and leaders haven't even solved for, as I was saying, like where we are today.
[00:20:10] Yeah. So if you look at your own company, you know, if you work at a big enterprise or something, you know, they're just still trying to get copilot to people and like figure out how to do it safely. And they're giving you these like neutered versions of it and stuff. Like that's the reality for most people.
[00:20:24] Most people aren't living on the edge of this capability. But this is why when I do my state of AI for business keynotes, I always talk about the dimensions of progress. And I try and like, show capabilities today, you know, show some examples for people, but then you lay out like, but here's where it's going.
[00:20:42] Like, all of this is just the foundation. So I talk about things like AGI agentic capabilities, getting more autonomous, more reliable, continual learning increases in memory. Like if you're using these tools every day, you're, you're, you've seen in the last few months, you know, turn on memory, like let it remember the conversations you're having.
[00:20:58] reasoning [00:21:00] capabilities keep getting better, recursive self-improvement, which is actually one of the areas that, that I think Anthropic is very concerned about is the better these models get, the more likely we are heading toward a path where they can improve themselves. And I think we're already starting to see that.
[00:21:15] And then world models is another one. So there's, I usually go through about, I don't know, there's like 12 or 15 dimensions, but those are some of the most common ones. So this then leads me to, this is a prelude to automated r and d and recursive self-improvement. So we know the labs are working on, you know, automating r and d within the AI models.
[00:21:33] something that should be very concerning to everyone is, while they're withholding this full release, this likely means that we're only nine to 12 months away from an open source model, being able to do the same thing.
[00:21:47] Mike Kaput: Yeah.
[00:21:47] Paul Roetzer: And then what, so like in essence, we have this very short window for all the banks.
[00:21:54] I mean, literally like every piece of software. Cryptocurrency. Like all of these [00:22:00] things in essence have to solve for this threat within the next nine months. Like, 'cause someone's gonna build this and release this. one of the other thoughts I had was, what would the other labs have done? Like mm-hmm. If Xai got there first, would they have the same restraint?
[00:22:17] one positive, I guess here is Elon did tweet over the weekend. Someone asked about like the, his promise of like more powerful models and he said it will take until May to be close to Opus 4.6 and then June to match or maybe exceed. So short time by normal standards, but long time in I arena what he's saying is like, Hey, we're not even up to Opus 4.6 yet, but like, we're working hard.
[00:22:40] So, you know, they're a little bit behind. one other topic that came to my mind is the government is continuing to attack Anthropic, their supply chain risk. And yet. They may be the only hope we have to protect our systems, our, our infrastructure, the software companies that we build around privacy of citizens like Anthropics at the [00:23:00] forefront of this.
[00:23:00] They're the only ones that are doing this and talking about this publicly in this way. and yet the government's treating them as the enemy. That's weird. concerns around the biggest companies having access to the future frontier models and then the potential centralization of power. So if you get into this situation where we get these massive models and they're so dangerous to release publicly that we only give them to Apple and Amazon and yeah, the banks and like, okay, well now we just centralized power.
[00:23:28] There's the broader implications on the security of all software, cryptocurrency, the ability to scale fraud on consumers and businesses. Mike, like,
[00:23:35] Mike Kaput: yep.
[00:23:36] Paul Roetzer: That one I think about like I, the amount of scams and spam that we're seeing and that like I'm sure are some in some way AI assisted for sure. Um. But if you give this kind of power to just the average scammer or the government actor that, you know, wants to destabilize things, like that's terrifying of, and I know, sorry, I'm just kind of like rambling here, but like, these are just the thoughts.
[00:23:59] So [00:24:00] another one is, use caution as an organization. so whether you're a team within a bigger company or if you're a startup, like an AI native startup, use caution when you're racing to integrate these age agentic systems into your organization. Mm-hmm. Just because Claude Cowork is amazing and open claw is fascinating.
[00:24:19] Like you have to remember how early this is and the tech is moving really fast, and even the people building it don't fully understand all the risks associated with it. So again, this is where I would caution, like on the bigger enterprise side, if it or legal is slow playing this stuff, that is, that is a good thing.
[00:24:40] Like I am. I totally understand the impact agents can have and how it can make your company have this massive competitive advantage. But I've yet to meet somebody who understands the risks of what they're doing when they're, when doing these things. So yeah, that's something we got. the compute and energy needs over the next decade may end up being dramatically [00:25:00] underestimated and underbuilt.
[00:25:01] So as crazy as it is that like, yeah, Google's spending 180 billion in CapEx this year. you know, like, you know, we're gonna have a trillion dollar, $2 trillion xAI pub, IPO, you're gonna have a o open IPO Anthropic. IP like I, my guess is we have like completely underestimated how much intelligence is needed and then the one positive I have here is this idea of project glass wing, that it does demonstrate the ability for like, the labs to work together.
[00:25:31] And I think that's gonna become, much more critical and then. There's just two other, thoughts I have. One is I would suggest people go back and listen to episode 1 41 again. So if you didn't listen to the Road to AGI and Beyond, I would go listen to that. it's an episode I did where I kind of walked through what I thought was gonna happen with a timeline of things were gonna be, and the two key components I wanted to just touch on is this idea of what accelerates progress and then what slows things down.[00:26:00]
[00:26:00] So what we're seeing is the acceleration through things like algorithmic breakthroughs, compute efficiency, large scale government funding, where they're like, now the government's getting involved, infrastructure investments, more compute capacity. Those are the things that allow it to go faster. But the things that slow AI progress down, and this is where I think mythos may be the preview of sort of what's gonna happen.
[00:26:21] failures and aligning AI models with human values, intentions, goals, and interests. That's what they're alluding to, is like we're, we're getting it more aligned, but like where it is misaligned is becoming much bigger problem. one of the other areas that could slow it down is restrictive laws and regulations.
[00:26:39] So heavy regulation of open source models. This mythos will likely accelerate this at a state level, so you're gonna see more bills being pushed forward to try and restrict this stuff because the federal government isn't gonna do it. And then the other thing you could see is, if there's a change in power in the [00:27:00] midterm elections in the US then, not the executive branch, but like, you know, at, at the, the house and the Senate, then we could see massive disruption, massive, issues where the Democrats will focus very, very heavily on regulation.
[00:27:19] They're gonna try and push this. And so that, that then is tied to this idea of societal revolt against ai due to job loss politics, perceptions, fears. And that is absolutely picking up steam. pushback on data centers is becoming very strong within some communities. Politicians are looking for wedges around like job loss and environmental impact.
[00:27:38] You're gonna touch, I think in the next topic, Mike, about, you know, what happened to Sam Altman like you're getting now. Yeah. Out, you know, people are out. If you didn't hear about it, somebody threw a Molotov cocktail at Sam Altman's house and then like 48 hours later shot up his house in San Francisco.
[00:27:54] So you're now getting people acting out against this stuff, which is insane and never the [00:28:00] answer. so you're just starting to see this. And then that leads to one of the other items that I'd highlighted in the what slows it down, which is voluntary or involuntary halt on model advancements due to catastrophic risks.
[00:28:11] Mike Kaput: Hmm.
[00:28:11] Paul Roetzer: That may end up being the most important one. So yeah, I, you know, I think there's so much more we could talk about on this one. I'll end with one other quick thought, and I think you've got this in the rapid fire, Mike, so I'm just gonna touch on it, but. Anthropic also released this emotions paper and it was about these models simulating or emulating human emotion.
[00:28:36] And I think it's a, something people should read. I'll just read two excerpts. One is, it said it may then be natural for these models to develop internal machinery that emulates aspects of human psychology like emotions. If so, this could have profound implications for how we build AI systems and ensure they behave reliably.
[00:28:56] And then Anthropic noted in this paper that none of this tells us whether language [00:29:00] models actually feel anything or have subjective experiences, but our key finding is that these representations are functional and that they influence the model's behaviors in ways that matter. So the reason I wanted to like include that in this commentary is we're looking at these like broad, far-reaching implications of these models.
[00:29:18] And in some ways it's kind of abstract to like wrap your mind around the significance of what's happening. And then when you come to this idea of like, but they're also showing signs of emulating human emotion. Mm-hmm. And so if you have these powerful models that can improve themselves, that can escape these hand boxes, that can identify zero days, which are, you know, unknown bugs within software systems, but they also have the ability to emulate human emotion, the ability to manipulate human emotion.
[00:29:49] We are, we are talking about like a perfect storm of, of future that we're just not prepared for. And to go back to [00:30:00] my original comment, why I think this may be a bigger deal than others, it's not that the mythos model is necessarily groundbreaking and we weren't aware that models were gonna get smarter.
[00:30:10] It's more about the moment where it might be what was needed for other people who aren't in the AI bubble. To be like, wait, what is AI capable of doing?
[00:30:22] Mike Kaput: Yeah.
[00:30:22] Paul Roetzer: and so maybe it starts these conversations down a path we really needed to be going,
[00:30:28] Mike Kaput: you know, and in the shorter term, I couldn't help thinking multiple times reading through all of this and the articles, if I am a, your average corporate IT person in charge of figuring this out, I just want to cry.
[00:30:43] Paul Roetzer: Oh yeah. Like, I just like, here you can have your ChatGPT licenses or whatever you want. And like the agent stuff just stay away. Like we're I,
[00:30:54] Mike Kaput: and even at best, if you somehow nail it, there's still going to be open source models nine [00:31:00] months from now that people are gonna use to bombard your company with cyber to
[00:31:03] Paul Roetzer: honest.
[00:31:04] Yeah. And I, I, the cyber stuff is, again, like I, you know, back in our agency days we had clients in cybersecurity that all these former FBI people working there and. There was people, you know, on our team that were working on those accounts, and I would just honestly be like, just, just tell me what I have to know.
[00:31:20] Like I, there's so much about cybersecurity, I don't, I don't want to know. and I, you know, even like even going through this stuff, your mind just starts to slip into like, oh my God. Like mm-hmm. How, how much they're going, the, like the bad actors are going to use this stuff is, we're just not ready as an industry, as a business world, as society.
[00:31:44] Like that is it, it, it, I think it's always been in the back. My mind is, one of the things I'm worried about. It is very quickly, like moving to the top of my mind of the things that I just, I don't know how we solve it. I'm not really sure how, how we figure [00:32:00] this out in the short time we have.
[00:32:03] Mike Kaput: Well, somewhat related in our next topic, Anthropic themselves as having a tough time figuring this out because they've also had another kind of high profile security incident because, in late March, March 31st, they accidentally leaked the entire source code of cloud code, which is their popular AI coding tool.
[00:32:23] This happened through a JavaScript source map file that was bundled into a public package. This file contained over half a million lines of unop guided type script across nearly 2000 files. So within hours, this code was downloaded, mirrored to GitHub and forked tens of thousands of times. Boris Cherny, the creator of Claude Code, said that basically their deploy process has a few manual steps and humans didn't do one of the steps correctly.
[00:32:50] So this was not AI related. Anthropic kind of flubbed a bit the response as well because they started issuing take down notices, for [00:33:00] thousands of GitHub repositories. But they were accidentally trying to knock down as well. Legitimate forks of Anthropics own publicly released Claude Code Repo.
[00:33:10] Cherney said they later retracted the bulk of the takedowns. this was also just immediately followed by Anthropic making a to some controversial move related to their subscriptions. journey also announced that starting immediately clawed subscriptions will no longer cover usage on third party tools like Open Cloud.
[00:33:30] Peter Steinberger, creator of OpenCL, called this move sad for the ecosystem, but gave churn credit for how he handled the communication. So Paul and is dealing with the consequences of their explosive growth and the popularity of Claude Code basically in real time. What did the last couple weeks here tell you about where they're at at a, as a company, like what challenges they're dealing with?
[00:33:54] Clearly there are a few.
[00:33:56] Paul Roetzer: the rate at which Anthropic has been shipping updates is, [00:34:00] I don't know that we've ever seen anything like it in business history. Never. Like they are just running circles around Google and openAI's, and everybody right now, it's, it's really remarkable actually. So the idea that like their systems aren't keeping up and the internal checks and balances, like, I get it.
[00:34:18] Like I don't, I just don't know we've ever seen a company grow this fast. Like
[00:34:22] Mike Kaput: no.
[00:34:23] Paul Roetzer: Their, their run rate right now is actually surpassing openAI's. Yeah. Based on reports from last week. They're like a $30 billion annual runway, which six months ago, if you would've said Anthropic, may IPO at a higher, you know, value than openAI's.
[00:34:38] I don't, I don't think too many people would've taken that bet. But if you, I don't know, there's probably market, there's probably odds on this right now,
[00:34:45] Mike Kaput: Rob. Yeah.
[00:34:45] Paul Roetzer: My instinct right now would be Anthropic will be a more valuable company than openAI's when they IPO and more valuable than X AI potentially.
[00:34:53] yeah, they're just, it's an incredible pace right now. What they're doing. The significance of the leak was one of the [00:35:00] questions I was thinking about. It's like, well, does this really matter? Like, they don't seem to air too much. I don't know. They just kind of keep moving and releasing all these other things.
[00:35:09] So the couple things that came to mind for me is it likely speeds up copycat models. So it, it made it easier for other people to sort of replicate what they're doing. It'll likely accelerate open source innovation because people can kind of look at this and it's not great for like, what we were just talking about with bad actors using these capabilities to do bad things.
[00:35:28] Mm-hmm. Like that. So those kind of jump out. The one I I will say is, I thought Boris was a amazing, like, as someone, you know, who comes from a PR and communications background, what he's doing is like just textbook stuff. And it, I think it's just totally organic and self-directed. Like, I don't, I don't think Anthropic was like, Hey Boris, like, go be the face of this problem.
[00:35:52] He just seems to be doing it and it, it's really impressive. So the way I'm watching it happen is his replies on [00:36:00] X or he is just interacting with people. So a couple of quick examples. someone posted like, 'cause obviously like a lot of developers are just drilling into this code, like what is it going on?
[00:36:09] What's in there? Mm-hmm. And so someone said, Claude code has a, a RegX. A RegX. Is that
[00:36:15] Mike Kaput: RegX? I think
[00:36:16] Paul Roetzer: re that detects WTF ff's piece of shit. FU this sucks, et cetera. It doesn't change behavior, it just silently logs is negative. True to analytics. Meaning when someone is working with Claude Code, the end user and they're like, this sucks.
[00:36:33] Like, or, or f you Claude code, like this is not good. Anthropic logs that reaction as a negative thing, but it doesn't change the behavior of the model. And so this guy who posted this was like, do this information what you will. Well, Boris responds and he said, this is one of the signals we use to figure out if people are having a good experience.
[00:36:54] We put it on a dashboard and call it the F'S chart. and so it's like [00:37:00] that. So it, they probably didn't really want people knowing that that was a thing, but rather than like saying, you know, like, oh, that's not, you know, we don't actually use that code or whatever. He's just like, yeah, it is what it is.
[00:37:13] then there was the other one. People are immediately like, oh my God, somebody's getting fired over this. So he, he has stayed really strong in this. He said it was human error. Our deploy process has a few manual steps and we didn't do one of the steps correctly. We have landed a few improvements and are digging in to add more sanity checks.
[00:37:30] Like with any other incident, the counterintuitive answer is to solve the problem by finding ways to go faster rather than introducing more process, in this case, more automation and Claude checking the results. And then he said no one was fired. It was an honest mistake. It happens. Then there was one other one I'll highlight that I thought was fascinating.
[00:37:49] So a user digging into the code, posts this OnX. He said, I can't believe more people aren't talking about this part of the Claude Code leak. There's a hidden feature in the source code [00:38:00] called Kairos, and it basically shows you Anthropics end game. Kairos is always on proactive Claude that does things without you asking it to.
[00:38:09] It runs in the background 24 7 while you work or sleep. Anthropic hasn't turned it on to the public yet, but the code is fully built. Here's how it works. Every few seconds, Kairos gets a heartbeat. Basically a prompt that says, quote, anything worth doing right now. it looks at what's happening and makes a call.
[00:38:28] Do something or stay quiet. If it acts, it can fix errors in your code, respond to messages, update files, run tasks. Basically anything Claude Code can already do just without you telling it to do it. But here's what makes Kairos different from regular code. It has at least three exclusive tools that regular code, Claude Code doesn't get.
[00:38:48] One push notification so it can reach you on your phone or desktop even when you're in, not in the terminal two file delivery. So it can send you things it created without you asking for them. And [00:39:00] three, pull request subscription so it can watch your GitHub and react to code changes on its own. Regular cloud code can only talk to you when you talk to it.
[00:39:09] Kairos can tap you on the shoulder and it keeps daily logs of everything. What it noticed, what it decided, what it did at night. It runs something. The code literally calls Auto Dream where it consolidates what it learned during the day and reorganizes its memory while you sleep. And it persists across sessions.
[00:39:27] Close your laptop Friday, open it Monday. It's been working the whole time, endless use cases. It's essentially a co-founder who never sleeps. The code base has this fully built and gated behind internal feature flags called Proactive and kairos. I think this is basically, or probably the clearest signal yet of where all AI tools are going.
[00:39:49] We are heading into the post prompting era where the AI just works for you in the background, like an all knowing teammate who notices and handles everything before you even think to ask. [00:40:00] This is absolutely what the labs are trying to build. So one, I mean, kudos. I don't Who was the guy who posted this?
[00:40:05] Mike, what was the username? I'd have to look.
[00:40:08] Mike Kaput: It was, yeah, we'll post it in the show notes. But yeah, the, I will also say if anyone from Anthropic is listening by any chance I, I'll pay a thousand dollars a month for this tomorrow. So need
[00:40:18] Paul Roetzer: to, and Boris's response. So again, he could just ignore this and just like let it go and not give it, you know, any fuel.
[00:40:25] He said we're always experimenting with new ideas. 90% don't ship because we don't think they're good enough. Experiences still on the fence about this one. Should we ship it? So he is just like, yeah, it's in there. You're right, we You got it. We, we built it and it's,
[00:40:41] Mike Kaput: they're on the fence about that one because of the compute problem.
[00:40:44] Paul Roetzer: Correct.
[00:40:44] Mike Kaput: Not, not the value of it. Sure.
[00:40:46] Paul Roetzer: Not on the fence enough to have not put it into the code that's already out there. Right. Meaning they're probably already using this internally.
[00:40:53] Mike Kaput: Mm.
[00:40:54] Paul Roetzer: yeah, so just fascinating stuff. And then the final note was just on the open claw impact. [00:41:00] And it kind of goes back to what I was saying earlier, like it's just a cautionary tale for companies that are out on the e edges here, that are building on the frontiers of the technological capabilities and relying on an unstable and infant AI ecosystem.
[00:41:12] So you know, it's, you're building an AI native company open clause, like, oh, this is amazing. We're all in like 30 days later, you've automated all these things and it's costing you like $2,000 a month.
[00:41:22] Mike Kaput: Yeah.
[00:41:23] Paul Roetzer: And then Anthropics like, yeah, no, that's misuse of the system and you just shut down your company into like today, or to do what you were doing is not gonna cost you a hundred thousand dollars a month, basically.
[00:41:34] So we just have to accept these like challenges and unknowns of building agents into workflows. And org charts is so early. So when you hear these stories of people doing it and you're so envious that they've figured something out that you haven't figured out, like they could wake up tomorrow and the thing they figured out is basically shot or it's like.
[00:41:52] So that, that's my main thing there is just so, so early.
[00:41:56] Mike Kaput: Yeah. I'll be so curious to see how that plays out. I don't know how some of these people are [00:42:00] affording to run these open claw setups on their own. Like just as a hobby thing, because I even hit some random usage limits in Quad Code over the weekend and was just like, oh, I've got hundreds of dollars of credits they gave me for various things over the year.
[00:42:16] and I was like, great, well we'll dip into the usage and in like four seconds I evaporated $300 on a random research check and I was like, how is anyone doing this dollar by dollar for every single thing you're doing?
[00:42:30] Paul Roetzer: Which we'll talk a little bit about the outcome-based pricing stuff in a minute.
[00:42:33] Mike Kaput: Exactly.
[00:42:34] Paul Roetzer: Just wild.
[00:42:35] Mike Kaput: Okay.
[00:42:35] Mike Kaput: Alright, our third big topic this week. There is a ton that has been going on with openAI's over the past couple weeks, so we are just going to go through some of these huge updates. Some good, some very bad, but. First up, OpenAI closed $122 billion funding round, which is the largest in Silicon Valley history.
[00:42:54] at the same time, Bloomberg is reporting that demand for OpenAI shares is sinking on [00:43:00] secondary markets, and the information reports CEO Sam Altman and his CFO are diverging a bit on IPO timing. It sounds like Altman wants to try to go public faster, whereas CFO Sarah Friar wants to maybe push it out a little bit due to spending commitments and the necessary organizational prep.
[00:43:18] Second, OpenAI acquired TBPN Daily Tech New show hosted by John Cogan and Jordy Hayes. This has become this hugely watched popular program in tech media. the show has only about 58,000 YouTube subscribers, but generated 5 million in ad revenue in 2025. They're on track to exceed 30 million this year.
[00:43:38] It will be housed within openAI's Strategy Organization. openAI's says the show will maintain editorial independence and continue choosing its own guests. Altman posted on X, TPBN is my favorite tech show.We want them to keep that going and for them to do what they do so well. Third, at the same time, a [00:44:00] major executive shakeup has hit the company Fidji Simo, the CEO of applications announced she is taking medical leave.
[00:44:07] She said a relapse of postural orthostatic tachycardia syndrome. P-O-T-S-A chronic neuro immune condition. She has talked about in public, quite a bit before she said to employees, she pushed, she's pushed a little too far and needs to try new interventions to stabilize her health. So there's some reshuffles related to this.
[00:44:26] President Greg Brockman will oversee product in her absence. COO Bright Brad Lightcap is moving to a new role focused on quote unquote special projects. And Marketing Chief Kate Rouch announced she is stepping down to focus on her recovery from late stage breast cancer, which she was diagnosed with a year and a half ago.
[00:44:44] Couple other things. Fourth, the New Yorker published a lengthy investigation by pretty famous journalists, Ronan Pharaoh, Andrew Marantz titled, moment of Truth, Sam Altman May Control our Future. Can he Be Trusted? This piece drew from over a hundred interviews and internal [00:45:00] documents, including ia, SVEs, slack messages, and Dario ADE's personal notes.
[00:45:05] And it basically builds this case that OpenAI systematically abandoned its safety first founding mission as its scaled up, and that Altman repeatedly chose to deprioritize safety commitments. And in fact, a former board member told magazine, he is unconstrained by truth. Now finally, we alluded to this days after this profile published, someone did throw a Molotov cocktail at Altman's San Francisco home.
[00:45:31] No one was hurt an hour later, police were responding to a man threatening arson at OpenAI's headquarters. Second attack in Altman's home followed. A couple days later, Waltman linked the attacks to the climate of AI anxiety and the negative media coverage he had eaten. Even written that someone had warned him the New Yorker peace came during heightened anxiety about AI making his situation more dangerous, and he responded to these attacks and the profile in a personal blog post, sharing a rare [00:46:00] family photo of himself, his husband and their child.
[00:46:03] he said he was sharing this in the hope it might dissuade the next person from targeting his home in a post. He acknowledged his mistakes and said he has this conflict, diversion that has caused organizational pain. And also concurrently Altman slash OpenAI went on kind of a major policy offensive.
[00:46:21] They published industrial policy for the Intelligence Age, a 13 page paper proposing a suite of people first policy ideas. Including giving every American citizen a direct state and stake in AI driven economic growth through a nationally managed fund seeded in part by AI companies. Vanity Fair reported.
[00:46:40] They're basically preparing a broader push to quote, rethink the social contract Axios framed this, framed this as Sam's Super Intelligence New Deal. So Paul, I don't know where to start. Lots going on here. Some of it really interesting, some of it very horrifying. Unfortunately, it's been a big couple [00:47:00] weeks.
[00:47:00] Paul Roetzer: Yeah, I'll, there's a lot of different directions. Go. I'll focus on Sam's post.
[00:47:06] Mike Kaput: Yeah.
[00:47:06] Paul Roetzer: and then the, the policy ideas. So one quick note. The TBPN, there's no confirmed. What did they pay for it? 'cause that's always, everybody obviously wants to know, but it does seem like it was north of a hundred million, which isn't bad for, you know, relatively new.
[00:47:23] Mike Kaput: Right.
[00:47:24] Paul Roetzer: the editorial independence thing, good luck. Like, I don't, I don't know these guys. I've actually never watched the show or listened to the show. I've heard of it plenty, but it's not something that's like, you know, intensely on our radar. but that idea of remaining independence as a media entity that's owned by an AI lab that has lots of pressures on it, that's gonna be very, very hard to, to maintain.
[00:47:47] But, you know, it sounds, I mean, they're gonna make their efforts too, so we'll see. okay, so then on Sam's post, I thought there's a lot of interesting things in here. So I, you know, the first, obviously the very personal stuff, [00:48:00] as I alluded to earlier, like violence is just never gonna be the answer here.
[00:48:05] And I do worry about these AI leaders. but it was only kind of a matter of time before something like this started to happen. in his post he said words have power. There was an incendiary article about me a few days ago, which is referring back to the New York article, Mike, that you just touched on.
[00:48:21] Mike Kaput: Yeah.
[00:48:22] Paul Roetzer: He said, someone said to me yesterday, they, they thought it was coming at a time of great anxiety about AI and that it made things more dangerous for me. I brushed it aside. Now, he did, later tweet that he sort of regretted the incendiary article reference and that, you know, he wasn't trying to pass blame.
[00:48:37] But, yeah, he did at least address that article. So then I highlight a few excerpts here on what he believes, and then he has some personal reflections and then his thoughts on the industry. 'cause his thoughts on the industry actually lead into the industrial policy for the Intelligence Age document.
[00:48:53] So on what he believes, he says, working towards prosperity for everyone, empowering all people, and advancing science and [00:49:00] technology are moral obligations. For me, AI will be the most powerful tool for expanding human capability and potential that anyone has ever seen. Demand for this tool will be essentially uncapped and people will do incredible things with it.
[00:49:14] The world deserves huge amounts of AI and we must figure out how to make it happen. it will not go all well, or all go well. He said The fear and anxiety about AI is justified. We are in the process of witnessing, witnessing the largest change to society in a long time, and perhaps ever we have to get safety right, which is not just about aligning a model.
[00:49:37] We urgently need a society-wide response to be resilient to new threats. This includes things like new policy to help navigate through difficult economic transition in order to get to a much better future. He also said AI has to be democratized. Power cannot be too concentrated. Control of the future belongs to all people and their institutions.
[00:49:56] AI needs to empower people individually, and we need to make decisions about our [00:50:00] future and the new rules collectively. And he said, adaptability is critical. We are learning about something, new very quickly. Some of our beliefs will be right and some will be wrong, and sometimes we will need to change our mind quickly as the technology develops and society evolves On the personal reflections, that this was kind of interesting he said, and again, I think in some ways he's actually like probably acknowledging some of the stuff from the New Yorker piece.
[00:50:24] Yeah. and other things that have been said about him. So I'm not proud of handling my, myself badly in a conflict with our previous board that led to a huge mess for the company. I have made many other mistakes throughout the insane trajectory of openAI's. I'm a flawed person in the center of an exceptionally complex situation, trying to get a little better each year, always working for the mission.
[00:50:46] We knew going into this how huge the stakes of AI were and that personal disagreements between well-meaning people I cared about would be amplified greatly. But it's another thing to live through these bitter conflicts and often have to arbitrate them and the [00:51:00] costs have been serious. I'm sorry to people I've hurt and I wish I had learned faster.
[00:51:04] And then on the industry, which leads into the the policy piece. So my personal takeaway from the last several years and take on why there has been so much Shakespearean drama between the companies in our field comes down to this. Once you see AGI, you can't unsee it. Yeah. It has a real ring of power dynamic to it and makes people do crazy things.
[00:51:26] I don't mean that AGI is the ring itself, but instead the totalizing philosophy of being the one to control AGI. The only solution I can come up with is to orient towards sharing the technology with people broadly and for no one to have the ring. The two obvious ways to do this are individual empowerment and making sure democratic systems stay in control.
[00:51:48] Laws and norms are going to change, but we have to work within the democratic process. Even though it'll be messy and slower than we'd like, I empathize with anti-technology sentiments, and [00:52:00] clearly technology isn't always good for everyone, but overall, I believe technological progress can make the future unbelievably good for your family and mine.
[00:52:09] While we have that debate, we should deescalate the rhetoric and tactics and try to have fewer explosions in fewer homes, figuratively and literally. And then that leads to the policy piece, which I would actually really recommend people read. It's only 13 pages. It, it's, it's a pretty quick read. I, I'll give you like a, a high level of what's, what's in there.
[00:52:28] So it's starts off within just a few years, AI has progressed from systems capable of fast, narrow tasks to models that can perform general tasks beyond, general tasks people used to need hours to do. Now we're beginning to transition towards super intelligence, which they say is AI systems capable of outperforming the smartest humans even when they are assisted by ai.
[00:52:50] No one knows exactly how this transition will all unfold. So then I'll just jump ahead to the two sections. In the paper they have building an open economy and building a [00:53:00] resilient society. So in building the open economy, they have, worker perspectives. So giving worker a voice in AI transition to make, work better and safer.
[00:53:10] They have AI first entrepreneurs help workers turn domain expertise into new companies by using AI to handle overhead that usually blocks entrepreneurship. They have, right to AI treat access to AI as foundational for participation in the modern economy. Similar to mass efforts to increase global literacy, modernize the tax base, AI reshapes work and production.
[00:53:34] The composition of economic activity may shift expanding corporate profits and capital gains, while potentially reducing reliance on labor income and payroll taxes. Another is Poten, public wealth fund. Create a public wealth fund that provides every citizen, including those not invested in financial markets, with a stake in AI driven economic growth, accelerate grid expansion.
[00:53:56] So establish new public private partnership models to finance and [00:54:00] accelerate the expansion of energy infrastructure required for power to power AI efficiency Dividends is an interesting one. Convert efficiency gains from AI into durable improvements in worker, benefits when routine workload declines and operating costs fall, including incentivizing companies to increase retirement matches or contributions.
[00:54:20] Cover a larger share of healthcare costs and subsidize child and elder care. adaptive safety nets that work for everyone. Make sure the existing safety net works, reliably, quickly, and at scale. 'cause if the transition to super intelligence is going to benefit everyone, the systems designed to provide economic and health security need to deliver without delay or gaps.
[00:54:41] Another is portable benefits over time, the public, or build benefit systems that are not tied to single employer by expanding access to healthcare, retirement savings and skills, training through portable accounts that follow individuals across jobs, industries, education programs, and entrepreneurial ventures.
[00:54:59] Two [00:55:00] more in this section, pathways into human-centered work expand opportunities in the care and connection economy, which they define as, childcare, elder care, education, healthcare, community services as pathways for workers displaced by ai. And then finally in that section, accelerate scientific discovery and scale the benefits.
[00:55:19] Build a distributed network of AI enabled laboratories to dramatically expand the capacity test and validate AI generated hypotheses at scale. And then the building a Resilient Society. There's a few here. Safety Systems for emerging risks. AI Trust Stack, which they say is research and develop systems that help people trust and verify AI systems.
[00:55:38] auditing regimes. So strengthen institutions such as a center for AI standards and innovation to develop auditing standards for frontier AI risks model containment playbooks, which we talked about, probably be pretty important as what we're seeing with Anthropic mission. Align corporate governance, guardrails for government use, mechanisms for public input, incident reporting, and [00:56:00] international information sharing in ai, around AI capabilities.
[00:56:03] So the other thought I have, Mike, and I'll just see if you have any thoughts on all this, but, maybe this is like my former PR background, but I'm thinking that the AI industry needs a massive PR campaign right now to highlight the potential for the positive changes in the world. And this like better future part of it is a PR campaign, but not in a way of like misleading people about what's possible and trying to like shift their focus from the negatives.
[00:56:32] That the negatives are real and they need to steer into those and not ignore them. Um. But what we need to do is accelerate some of the wind that have, positive impacts in society, that are high value, high profile, that could build excitement about a better future. Things like drug discovery and curing of diseases.
[00:56:50] And we know like they're working on these things, but I feel like right now the negative sentiment is just like snowballing. Yeah. You can feel it every week in the topics we're covering [00:57:00] and the articles we're reading, and there's very few really positive things. And so, they all, all the labs, they need to figure out a way to do this where they acknowledge the negatives and do what they're doing, but they, they gotta start getting some big wins or else society's gonna turn on this stuff fast.
[00:57:19] And I don't know how, how fast you can go on the scientific discovery, but I keep coming back to that is the thing that's, that's gonna change perceptions is if you can actually improve people's lives and vary clear ways, that, that. You're gonna need to win mind share, and right now they're losing.
[00:57:38] It is kind of my current take on the industry.
[00:57:42] Mike Kaput: I could not agree more. I'd love to, for us to even talk more about initiatives like that and maybe future episodes and work on that. Because I would also just encourage, you know, I'm by no means then expert on what you should be doing in terms of your messaging here, but it would also strike me as [00:58:00] valuable for especially Silicon Valley based AI labs, to also focus on the individual.
[00:58:05] How do these things make your individual life better? The big picture stuff, super important and really valuable, but also think about all the things that people are going to be upset about when it comes to an AI lab. They do not want you telling them that you're going to save the day, that their life is gonna be managed by your technology.
[00:58:24] Show them how it empowers them and how real people are using it for real wins. Even basic ones in their life, I think could be also interesting as an.
[00:58:33] Paul Roetzer: That's a great point. Yeah. And I do think, like you and I see, you know, a lot of the similar stuff. Like right now all you have is these individual stories on acts that never break out of acts, the x bubble.
[00:58:44] And it's like these incredible stories of finding cures for things that their doctors missed for years, finding treatment paths. And I've certainly experienced that myself, like things in your own personal life where you're just like, I don't know what to do. and you just like, have a conversation.
[00:58:59] It's like, wow, [00:59:00] okay, that, that, that's the direction. Like, I think I know what to do. and there's like, I'm sure there's just all those incredible stories, but right now, yeah, I just, I feel like they're just missing it.
[00:59:10] Mike Kaput: yeah.
[00:59:11] Paul Roetzer: Yeah. I don't know. I think you're right though. Like, we should, we should make a bigger effort on this show to like highlight more of that stuff.
[00:59:17] I think there are so many amazing things that are happening, especially on the scientific discovery side and, you know, making an impact on people's health and wellness and things. Mm-hmm. yeah. We should do more.
[00:59:30] Mike Kaput: All right, Paul, before we jump into rapid fire quick announcement, this episode is also brought to us by our AI for Writer Summit.
[00:59:38] So the future of storytelling is being ri rewritten thanks to ai, and that's why we're very excited to be hosting our annual AI for Writer Summit on Thursday, May 7th. So this is a half day virtual event for writers, editors, content teams. Basically anyone who does any type of writing or content creation as part of their work, you will get tons of [01:00:00] awesome actionable knowledge from the event because during it, we'll have some incredible speakers breaking down exactly how AI can help you create smarter and faster, but also importantly, without kind of losing the heart and soul of your writing.
[01:00:12] This event has a free registration option, so go check those out today. You can go to ai writers summit.com or just go to marketing ai institute.com and click on events and you'll find. The summit right there. And by the time you go to the website, the agenda will be live. So you can see the great lineup we've got going for you.
[01:00:32] Super excited for this one.
[01:00:34] Paul Roetzer: Yeah. And real quick note on that, I mean, last year we had, I think it was more than 4,200 people. Yeah. From 95 plus countries. So
[01:00:40] Mike Kaput: yeah,
[01:00:40] Paul Roetzer: it's an amazing event. It's a great way to network with other people. And then the real key is like, we're trying to tell the human side of this.
[01:00:46] So this is not like how do you automate the writing and get rid of people? We're trying to grapple with the hard questions, you know, like what is the future of journalism? What is the impact it has on people who write for a living, for fulfillment, things like that. So we very [01:01:00] much focus on that. And then I'm, if I'm not mistaken, like I think my opening keynote from last year might be on YouTube.
[01:01:06] If not, we'll put it up on YouTube. Yeah. Before tomorrow we'll put a link into it. So I did the state of AI for writers and creators navigating the future of creativity. But what I focused on last year was the human side of it. And when should we use AI to write was like the question I po posted? or challenged people with.
[01:01:23] And then I actually presented a framework to decide like, when should I use AI versus when should I not? And so I think it's a really important concept. So we'll put the keynote from last year up, that people can go and watch. and you know, I think it's a, a good way to get into like a 25 minute keynote, if I remember correctly.
[01:01:39] Mike Kaput: Yeah. Awesome.
[01:01:41] Mike Kaput: Alright, let's dive into some rapid fire in late March, AI recruiting startup Mercor was hit by a supply chain cyber attack through a tool called LiteLLM, which is a widely used open source library that connects applications to AI services. A hacking group claimed credit and published samples of the stolen data.
[01:02:00] TechCrunch reported. These included Slack messages, internal ticketing information and videos of conversations between Mercor Corps's AI systems and contractors. Now, the reason this matters, why we're talking about it, we have talked about Mercor before. They are a $10 billion company. That provides training data to the top AI labs.
[01:02:18] So what they do is they recruit expert contractors. So think people like engineers, lawyers, doctors, bankers, and they have them train AI models and chat bots. Some of their top customers include openAI's, Andro, and Meta. They have more than 30,000 experts on their roster and say that they are paying $1.5 million per day to their contractors.
[01:02:39] So there's a lot of data in this system and the attackers claim to have attained four terabytes of data in total, including source code and database records. Not only is this bad from a personal perspective, 40,000 contractors at least have had personal data exposed, it sounds like. they've also exposed proprietary source code video interviews, and the most [01:03:00] important part is potentially this could include details of how Frontier Labs are training their models.
[01:03:05] What kind of expert feedback they're collecting and the methodologies behind their most advanced system. So, so far Wired has reported that Meta paused its work with Mercor and is investigating the incident. openAI's confirmed it was investigating its exposure, but said it had not paused or ended its contracts at this time.
[01:03:23] So, Paul, another security incident. We've covered Meco in the past, how important it could be to the AI ecosystem, though this is a pretty damaging, series of events. We also actually did talk about that LiteLLM breach a couple weeks ago. So two topics kind of coming together in less than ideal ways.
[01:03:40] Paul Roetzer: Yeah, like I said, I hate talking about this stuff. It, I really do like it, it is terrifying. Yeah. and what we know is like the, like the, when, when state actors want something, they're gonna get it. Like diode did this interview back in like 2023 or 24. That just always haunted [01:04:00] me. Where he was talking about like, the weights to these models are literally like, these are like the nuclear codes basically, in terms of how they protect these things.
[01:04:08] There was actually an example recently where they were talking about openAI's, literally going in with like the briefcase, like the, what do they call it? The football, the nuclear football, whatever. The nuclear football. Yeah. Like that's how they delivered the model to like with the weights in this lock.
[01:04:20] Oh, my case to the government when they were like trying to build a custom version of something for the government. So the weights to these models are so tightly held. Yeah. like I think Dario said at the time, there's like two or three people within Anthropic that even had the ability to, to know the weights kind of thing.
[01:04:35] and he said like, listen, if a state actor wants to get 'em, it's just how much money are they willing to spend to go get 'em? Like they could hack into anything. And so the premise that like all, you think of all these areas of risk and all this data that's living in these companies and like this maybe partially goes to this use caution when you're like working with just.
[01:04:56] Random startups and giving them access to your APIs and like all this [01:05:00] shit, like, you're just, the surface area of risk is so vast and misunderstood or like misunderstood by people. Yeah. it really is just terrifying. I don't like, cybersecurity to me is just,
[01:05:15] but
[01:05:15] I, like I said, I think like cybersecurity professionals, lawyers who deal with this stuff, like, man, talk about safety and like, you know, what jobs to go into.
[01:05:24] Like
[01:05:25] Mike Kaput: I guess that's a good, a good silver lining here, right?
[01:05:28] Paul Roetzer: Yeah. We, we may make for all the lost jobs, everyone's just going into cybersecurity.
[01:05:32] Mike Kaput: Everyone's fixing all the new nightmares. AI is enabling.
[01:05:38] Paul Roetzer: yeah, but this is a bad one. This is a bad one. That's
[01:05:39] Mike Kaput: the, yeah. Yeah. And it's good to be aware too, of these companies like Mercor that, you know, in Silicon Valley Circle is definitely well known, but maybe to your average public, not as well known, or a household of a name, but super, super important to the ecosystem.
[01:05:52] Paul Roetzer: Yeah. And real similar, like, scale ai, right?
[01:05:55] Mike Kaput: Scale ai for sure. Yeah, yeah, yeah. In fact, there was something in the reporting where meta, [01:06:00] you know, when they, even when they essentially acquihire scale ai, they didn't stop using Mercor or either they were just using both because it was so important.
[01:06:08] Paul Roetzer: Yeah.
[01:06:09] Scale ai. If you don't catch the reference, a Alexander Wang who's now in charge of super intelligence at Meta, he was building a training company called Scale ai. He got acqui hired for like $15 billion by Meta. So his company still exists, but
[01:06:22] Mike Kaput: yeah,
[01:06:23] Paul Roetzer: that, that's, that's the reference there.
[01:06:25] Mike Kaput: All right, well, this next topic's a little more, positive or at least interesting and not negative, right?
[01:06:30] But in early we need something, right? Right. In, in early April, Andrej Karpathy posted on X about how he is now using LLMs, not just to generate code. You know, he is a programmer coder, so he is doing that a lot, but also to build and maintain personal knowledge, wikis. So this post, as of today, has nearly 20 million views, so it's like one of the more viral AI posts this year so far.
[01:06:54] And the core idea here is that instead of relying on all this technical stuff like vector databases and complex rag [01:07:00] pipelines, instead he's just dumping raw documents, articles, and research into a folder. Then letting an LLM compile them into a structured interlinked markdown wiki. And then he uses obsidian a free note-taking app as basically the front end of this.
[01:07:14] So as he puts it, obsidian is like the IDE. The LLM is the programmer, the wiki is the code base. So this LLM then handles, curating sources, linking updates, and even runs periodically to check for inconsistency. So the reason this is kind of getting some popularity in some eyeballs is because like every knowledge worker in some way is using kind of information and knowledge bases that are really often very hard to maintain.
[01:07:41] So instead of just thinking about LMS as Chatin interfaces or code generators, Karpathy is really thinking about this in terms of LLMs turning, becoming persistent knowledge infrastructure and building that out in ways that compound over time. And people kind of ran with this and started building their own versions.[01:08:00]
[01:08:00] Obsidian founder weighed in with best practices. and Paul, I just thought this quote from Karpathy was telling, he said, you know, in this way and the way I'm using this, a large fraction of my recent token throughput is going less into manipulating code and more into manipulating knowledge stored as markdown and images.
[01:08:16] Super interesting implications for, you know, maybe less technical people.
[01:08:20] Paul Roetzer: Yeah. The term we hear thrown around a lot in the last like 30 days was the idea of a second brain. Yeah. Like everybody's kind of talking about the, this idea, like, all your information just lives in this thing. And so, you know, the major cloud companies are trying to solve for this productivity.
[01:08:33] Companies like Microsoft and Google, obviously they want this to just, you already have a lot of this information living in there and they're trying to find ways to like, make it easier to build these sort of second brains where all this information lives there and the knowledge base is there. and then you're just constantly like, you know, almost like that idea we talked about with the Claude code leak, where just proactively acting on all this knowledge and just like working with you on it.
[01:08:54] Um. And the thing with carpathy posts is, you know, three months from now somebody will [01:09:00] productize what he's doing or maybe three days from. Sure. Yeah. so he talks in these technical ways and most people aren't able to do anything like what he's explaining. so the average business leader, or practitioner listens to our podcast.
[01:09:14] I don't know what any of that means. I don't know what Id ID is and things like that. but for everybody else, just like assume the outcome of the idea is like a product waiting to be built. And then that's like the premise here is if he's talking about it being possible, it's only a matter of time until someone like builds that capability and then you start finding it.
[01:09:34] Like all of a sudden you have access to that. You can kind of hack it together with the things you've got internally.
[01:09:39] Mike Kaput: Yeah. It struck me too, as related to another thing we had talked about that he was working on that auto researcher concept where it's like. I just was like making notes while reading through his post and saying like, this feels like in some fashion, whether it's doing it yourself or there's a product around it, that every analyst and research firm basically needs to go this direction [01:10:00] at some point because you need this second brain of all this proprietary.
[01:10:03] Stephanie. I know people are kind of doing it in layering chat over it, but this is dynamic. It is updating regularly. It is an LOM maintained wiki or knowledge base or second brain, and I think that's probably where I'd imagine research function should be going.
[01:10:18] Paul Roetzer: Yep.
[01:10:20] Mike Kaput: Alright, so next up. There's a lot that's been happening on the AI and jobs front, no surprise in the past couple weeks.
[01:10:27] So we're gonna run through a couple highlights here of some things that are notable. So first, the New York Times published a piece reporting that economists who had previously dismissed the AI job threat are now slowly but surely starting to change their minds. So this is a pretty. Big shift in establishment economic thinking.
[01:10:44] They talk to a bunch of economists who, you know, they're not doing a total 180, but they are starting to acknowledge that maybe this mainstream economist position that AI will create more jobs than it destroys the way previous waves of technology have. Maybe this is a little out of date or there's more [01:11:00] nuance to it than previously thought.
[01:11:02] Second, there may be data backing that up. The Challenger Report, which is a regular report we talk about for March, 2026, attracts job cuts. challenger and Gray is a, recruiting firm. They show that US employers announced just over 60,000 job cuts. March, 2026, that's up 25% from February. AI was cited as the leading reason for 15,000 of those cuts.
[01:11:28] So it's about 25% of the total year. To date, AI ranks a fifth among all the reasons for job cuts. And since Challenger began tracking AI as a layoff reason in 2023, the cum cumulative total, so of all time has now crossed 99,000 AI related job cut announcements across three years. Third, Jack Dorsey we talked about a couple weeks ago, is making the case for AI driven restructuring much more explicitly perhaps, than any other major CEO out there.
[01:11:58] So after block cut, [01:12:00] 4,000 of its more, more than 10,000 employees. Dorsey has now published a blog post co-written with a partner at Sequoia Capital, arguing that AI should replace the entire traditional hierarchy of middle management. So block, he says, is restructuring around three employee roles.
[01:12:16] Individual contributors who build systems is one. Two is directly responsible individuals who own specific outcomes on 90 day cycles. And third is what they call player coaches who mentor while staying hands-on with technical work. He said this restructuring was triggered. By a capability shift he observed in December with Anthropics Opus 4.6, and OpenAI's Codex 5.3 fourth.
[01:12:41] On the hiring side, Zapier released the second version of its AI fluency rubric, which now they apply to every new hire at their company. This requires candidates to demonstrate AI embedded into their core work, not just one-off usage. They wanna show repeatable systems and measurable impact on quality, efficiency, or outcomes.
[01:12:59] [01:13:00] They also have this new accountability dimension that they consider. They say with AI you can delegate the work, but not the accountability. So keeping that human in the loop hide top of mind here. Zapier's language is also pretty blunt about their AI expectations. They say if someone isn't meaningfully improving their work with AI support, they just don't meet the bar.
[01:13:19] And then last but not least, a new Gallup survey shows that AI is reshaping how college students think about their futures. 42% of bachelor degree students surveyed, so they have reconsidered their major because of ai. 16% said they've already changed their major over it. For people trying to get associate degrees, 56% are also reconsidering their field of study due to what AI enables.
[01:13:41] So Paul, what, what jumped out to you about these updates this week? I mean, I'm personally planning on diving in a lot deeper to Dorsey's thoughts. I thought this were kind of interesting. I
[01:13:50] Paul Roetzer: read, I read a i that might have been the thing that triggered, so I put a post on LinkedIn on like, I don't know what day it was.
[01:13:55] It was, it was one of the days we were in Scotland and I was like, we were driving a long distance and we were sleeping [01:14:00] in the car and Cat, I was like typing away. We had a tour. I was not typing while I was driving. I, we actually had like a tour guide driving us. I don't remember which thing I had read that I wrote the LinkedIn post about, and then I turn it into a newsletter post.
[01:14:12] it might have been the Dorsey one, I don't remember, but it, it was abstract. Like his, yeah, I read it. 'cause I'm very interested in this. I'm actually like, my Mayon keynote this year is gonna be based on like a vision for AI forward org chart. I think, maybe like. I, I'm working through an idea, Mike, you've seen some early versions of this.
[01:14:30] so I'm very, very keen on this idea of organizational structure and what teams are gonna look like and things like that. So it did definitely catch my attention what they were doing. I love Zapier's approach. I liked it when they came with the V one. I really liked the V two. I liked the idea of the say I fluency rubric.
[01:14:45] So that was some really cool stuff. and then just the jobs overall, like, again, like, I'm glad to see people coming around and realizing like, this is a real thing and it's gonna be a problem. The thing I alluded to that I wrote about on LinkedIn though, was, [01:15:00] I'm getting really, really annoyed by the, like the tech leaders in particular who just keep pretending like it's all gonna be great.
[01:15:09] Yeah. Like with no acknowledgement of the possibility that it won't be. So I get optimism. Like I'm all, I'm all for being optimistic about this stuff. and like believing in a future of abundance and like, we're gonna find our way through, which I do think we will. Like, I think it's gonna end up being great.
[01:15:25] But I also like straight up, like it's gonna suck for a lot of people in the process. Like this isn't gonna be an easy transition and like a whole bunch of people are gonna lose their jobs. And, so I get really annoyed when people won't like, acknowledge both sides of the equation. So the example I I put in the newsletter was I said tech leaders, politicians, economists who point to increasing demand for software developers and historical precedents as proof that AI won't displace millions of jobs are creating a false sense of hope.
[01:15:52] And that I highlighted for in particular, and some of these people are people I respect and like follow, but Andreessen Mark Andreessen, this is a quote and we'll put the links in the show notes. [01:16:00] The AI job loss narrative are all fake. AI equals mass, massive ramp and productivity equals massive ramp and demand equals massive jobs.
[01:16:07] But watch, so that was a tweet from April 5th.
[01:16:10] Mike Kaput: Mm.
[01:16:10] Paul Roetzer: Aaron Levie, who we really like, like CEO of box, like I'm a big fan of Aaron. He's got some of the best takes on x about AI that I've seen. He does a lot of research on this topic, so. April 5th, he wrote, there are far more categories where AI agents making things more efficient will induce demand for that skill than spaces where agents eliminate the work.
[01:16:30] This is why the AI job predictions will not play out as advertised. okay. Shyam Sankar, who's the CTO at Palantir, we've talked a lot about Palantir. he had an editorial February 2nd. He said, AI is a tool for the American worker, not his replacement. The job loss narrative is a ploy to attract investors, drive media attention and consolidate political power.
[01:16:50] The real promise of AI is the enterprise, is to make the American worker 50 x more productive to unleash his taste and agency. This isn't [01:17:00] speculation, it's reality. It's very, very confident. There's lots of confidence in these statements. And then David Sachs, this is no surprise, he, he, he can't acknowledge the impact on jobs due to his relationship with the administration.
[01:17:11] He is the currently the chair of the President's Council of Advisors on science of technology. All caps, ai, job loss, hoax exposed. And then it goes on to say, according to a new study from Vanguard, the occupations most exposed AI automation are actually outperforming the rest of the job market in terms of growth and real wage increases rather than causing job loss, AI is making workers more productive, driving gains in both jobs and wages.
[01:17:35] So what I said was, despite these economic or optimistic outlooks from these leaders, the reality facing companies, especially those with limited growth and demand, which is a really important asterisk here
[01:17:44] Mike Kaput: mm-hmm.
[01:17:44] Paul Roetzer: is that the pressure to reduce headcount across all areas of knowledge work is going to be immense in the coming months and years across all areas, marketing, sales, customer service, hr, finance, et cetera.
[01:17:55] and then I said, pretending like there isn't at least a strong possibility of [01:18:00] significant disruption is a disservice to business leaders who should be doing more to prepare their organizations and upskill their people. And then I said, I talk to executives every week who are being told to stay flat on headcount and to have a contingency of cuts ready to go if the efficiency from AI happens.
[01:18:18] And so I just, I don't, that's like my continued frustration is that all these people who are hyping AI as this future of abundance, which I, I'm, I'm with you. Like, I hope, and I do think eventually, but I don't know who you're talking to that is planning to hire, like I'm not meeting those people, like, unless they're Anthropic or one of these companies that's growing at 20, 50, a hundred percent a year.
[01:18:43] I, I've yet to talk to an executive at a traditional enterprise that's really happy with five to 10% annual growth that's planning to hire, like there, it's not happening. So, and that's in knowledge work. Now, of course, there's exceptions to that in energy, in the trades in [01:19:00] healthcare. Like yeah, we can't hire enough people in those areas.
[01:19:02] I get that. I'm talking about the rest of us, all the other industries where the ultimate goal right now is to, to just stay flat and headcount and get the revenue per employee number way up. So, yeah, I don't know. It's, I guess it's good, like it's increasingly becoming a conversation because it just really needs to be, we need to be thinking about what if these tech leaders were so optimistic just are wrong?
[01:19:27] Yeah. And what if it isn't, as, you know, easy and of a transition as they'd like to make you think.
[01:19:34] Mike Kaput: So similarly, we've had a lot happening on the AI policy and politics front. So we're gonna go through a few developments here that have happened over the last couple weeks. So, first up, California Governor Gavin Newsom signed a first of its kind executive order requiring safety and privacy guardrails from AI companies that contract with the state.
[01:19:53] So this basically establishes new certification requirements for AI vendors that want to do business with California. It [01:20:00] requires them to attest to and explain their policies around preventing illegal content, harmful MO model bias and violations of civil rights. It also directs state agencies to expand the use of vetted AI tools in government.
[01:20:13] Developed an AI powered pilot for accessing government services and publish a data minimization toolkit. Second, at the federal level, the Wall Street Journal reports that the White House is racing to head off threats from powerful AI tools. There's renewed urgency here in the wake of the all the mytho stuff we discussed.
[01:20:31] this included prominently a group of White House officials working on the issue, including convening a call with the vice president, treasury secretary, and the heads of Anthropic, openAI's, Microsoft and Google, as well as the leaders of cybersecurity firms, CrowdStrike and Palo Alto Networks. That's obviously in addition to the previously mentioned meeting we talked about related to mythos, specifically that the Treasury Secretary had with bank CEOs.
[01:20:57] Third, a major new survey from Fathom. A [01:21:00] non-partisan research organization provides a clear picture of what Americans actually want from AI governance. I surveyed a bunch of people to ask about their feelings and priorities in a number of areas. So the top priorities across party lines for people are child safety, corporate accountability, and verifiable standards.
[01:21:17] Another big issue is workforce protection. So according to Fathom, from retraining programs to sovereign wealth funds that share AI generated wealth with the public, every workforce policy tested in this survey commanded majority support Americans decisively reject, leaving workforce transition to market forces.
[01:21:38] There's broad demand, but no preferred solution. A policy window that is open now but won't be indefinitely. And then lastly, Politico reports that Senator Bernie Sanders may be building an unlikely alliance with Silicon Valley AI safety advocates. So Sanders recently met with. Quote, unquote, AI doomers in Berkeley, including Eliezer Yudkowsky [01:22:00] from the Machine Intelligence Research Institute.
[01:22:03] And, Sanders said, I know there have been a lot of science fiction novels in movies about how the robots and the AI and the computers rebel against human control, but these guys no longer think this is science fiction. So, political Politico suggests this might be the beginning of an alliance between anti AI populists and the more tech-centric, perhaps effective altruist, aligned AI safety advocates.
[01:22:27] Paul, did anything jump out to you here? There's a, there's a wild quote from Yad Kowski and the Politico piece where he just said, basically telling Bernie Sanders, Hey, the point, if AI gets much, much, much more powerful, it'll run everything. And Sanders said, what does that mean? Humans are discarded.
[01:22:46] And yet Kowski replied, think everybody dead. So there's some, some strong language being used.
[01:22:53] Paul Roetzer: Well, I mean, just like the optimist side we just talked about. There's extreme views of [01:23:00] everything.
[01:23:00] Mike Kaput: Yeah.
[01:23:00] Paul Roetzer: I'll just, I'll probably just leave it at that at the moment. I, but this is my concern is like, does un you have an uneducated public largely about what these things really are, what the real risks are, what the real potential is.
[01:23:14] and so there's always, when you have, when the, when the literacy isn't high Yeah. And the comprehension isn't high, then you have the ability for extreme views to come in and influence people's perceptions and beliefs. And that's very dangerous in my opinion. you know, if you, if so, let's say, you know this Yudkowsky, is that how you say his name?
[01:23:41] Mike Kaput: I believe so, yeah. Yudkowsky. Yeah.
[01:23:42] Paul Roetzer: So like, let's say that's the first thing you hear. It's like this ends up on a 60 Minutes or in that AI movie that just came out. Yeah. The Domer AI movie that was out, um. Yeah. And you hear that and it's like, oh, I hate ai, I hate data centers. I hate Sam Altman. And like you people take that perspective and then you say, [01:24:00] yeah, but it, it's, it found a cure for cancer to your family member that, you know, was suffering from cancer.
[01:24:07] Like, it, AI is actually the thing that's gonna find the cure. Did Find the cure. Like, so should we stop it? Should we not, should we not have AI now? Because you, you heard a bad quote or like, so, so, and that's when there's like all these nuances to when the people take extreme views, they, they don't stop and then say, oh, okay, well maybe that would be amazing.
[01:24:29] Yeah. and again, I, the only pair I could ever go back to is the internet and then say like, oh yeah, the internet's gonna allow these scams and dark web and all these, like, horrible things are gonna happen, but it's also gonna open up the economy and we're gonna be able to build all these amazing things.
[01:24:44] You're gonna connect with people. You could never, and you're gonna be able to FaceTime with your family when you're a thousand miles away. You want to, you wanna not have any of that good stuff. You just wanna like, should we just shut it down? Because like
[01:24:54] Mike Kaput: right.
[01:24:54] Paul Roetzer: Bad stuff might happen,
[01:24:56] Mike Kaput: right?
[01:24:56] Paul Roetzer: You can't.
[01:24:57] and so that's like, my feeling on this is [01:25:00] like dialogue and reason and finding paths forward where we can do this responsibly, but this, this absurdity of like, shut it all down because it's going to eliminate everybody. It's like, okay, that's your belief. And then, then we don't get any of the good stuff either.
[01:25:18] So how about we actually like, just be reasonable here and find the reality and like, let's talk about the reality of the situation and not take extreme views that are unrealistic and mislead people. So.
[01:25:32] Mike Kaput: Alright, so next up, HubSpot announced that starting April 14th, it's customer agent and prospecting agent are moving to outcome-based pricing.
[01:25:41] So this means, according to HubSpot quote, customers only pay when the agents complete the task it's been assigned. Practically, this means the customer agent is moving from used a used to, or will used to have charged you $1 per conversation no matter what. And that's moving to 50 cents per resolved [01:26:00] conversation, meaning you only pay when the AI actually solves the customer's problem.
[01:26:04] The prospecting agent is shifting from a recurring monthly charge per enrolled contact to $1 per lead recommended for outreach. So HubSpot says quote, this means you now pay when a prospect's prospect gets qualified and handed to your team. HubSpot says, customer agent actually now resolves 65% of conversations and cuts resolution time by 39%, and that prospecting agent activations are up 57% quarter over quarter.
[01:26:32] Both agents include a free 28 day trial and are available to pro and enterprise customers. So Paul, we've talked about this before, the need for SaaS companies like HubSpot to update their pricing models. What do you think of this approach? Are they headed in the right direction?
[01:26:48] Paul Roetzer: I love HubSpot. I always have to preface my comment on HubSpot.
[01:26:52] I love 'em. of the people at HubSpot, I love the company. We use the technology like our company's built on the technology, right? Both my companies over [01:27:00] the last 20 years have been built on HubSpot technology. I like the concept of the outcome pricing. I understand from HubSpot's perspective why they would move to this.
[01:27:12] I understand the messaging of why this is a benefit to customers. So I get all of that. Mike, you and I have worked in HubSpot for a really long time.
[01:27:23] Mike Kaput: Yeah.
[01:27:24] Paul Roetzer: The reliability of the data is a problem. Yeah. allowing them to determine what is a resolved conversation is a very, very gray area.
[01:27:35] Mike Kaput: Yeah.
[01:27:36] Paul Roetzer: so we get a spam form fill or something like that, or a spam chat and like you close it and now we just paid 50 cents to close a spam chat.
[01:27:44] Like. There's all these, like, what does that mean? Like what is a resolved conversation? and now the work we have to do to go into understand what is a resolved conversation. Mm-hmm. Oh crap. Like 80% of what would be considered [01:28:00] resolved conversations to us are considered not evaluate at all. Like how do we turn that off?
[01:28:06] And, or $1 per lead recommendation for outreach. Like, I don't know. Like that doesn't, I, it doesn't, as the CEO EO of a company that pays lots of money to HubSpot every month for software, neither of those jump out to me right away is like high value things that I want to like now have to like talk to my COO today and be like, what does this mean?
[01:28:30] 'cause it goes into effect tomorrow and like, alright, have we budgeted for this? How do we budget for this? So, no, I, this creates way more questions for me than answers. Maybe it ends up being a really good play. I don't know. But it continues down this path of the way these companies are trying to price AI to me is, they're not, they're not solving this, what is the simplest answer to this?
[01:28:54] They're, they're, it just keeps getting more complicated in my opinion, even though it's now outcome based, it's still [01:29:00] uncertain to me.
[01:29:01] Mike Kaput: I always even just think back to even the agency days where it's like, Hey, we ran this really successful campaign for you. Look at all these great leads. And then they're like, yeah, that's awesome, but like, these leads aren't closing or these aren't qualified.
[01:29:13] And you're like, well, yeah, they are. And it's like, well, no, they're not because they're not closing. And then you find out there's a hundred other issues of why this wouldn't be the case. So I wonder how that you even get to that shared agreement on any of these outcomes.
[01:29:24] Paul Roetzer: Yeah, I, and again, like, we'll dig, we again, this affects us personally, so we have to dig into this.
[01:29:30] So we'll report back like again, maybe it is a really elegant solution. And maybe it is truly value-based for us as the customer. Unfortunately, my instinct is this is gonna be a pain in the ass and I'm not going to agree with the value that they're assigning to these. And we're gonna have to now figure out ways to either not use breeze for these things or to like change the dynamics or certainly update our budgeting.
[01:29:57] It's just super annoying and I [01:30:00] understand why they have to do it, but no, I'm not super excited about this. If they would've come back and said, we're raising your monthly seat license, $10 per month per user, and you have unlimited, I'd have said like, great, raise it 40 a month if you want to. I don't give a shit like, as long as I don't have to think about this and there's no additional budgeting, then just.
[01:30:21] Just raise the rate. If the software's creating that much value, then just charge me more. I don't know.
[01:30:26] Mike Kaput: Mm-hmm. Yeah. And again, we'll see. I don't wanna harp on it, but I'm just even thinking like, my God, we sell like 10 different things. Like I don't think all those leads are the same. You know, there's like, like multiple lines of business, all these crazy considerations that maybe our solve for.
[01:30:39] But I'm just like, oh, I have more questions than answers.
[01:30:42] Paul Roetzer: Yeah. My brain explodes with the questions and the reasons this won't work versus the, oh, thank God you solved the pain point for me. It's like, no, you just gave me 10 more.
[01:30:51] Mike Kaput: Right. Alright, a couple more final segments here before we wrap up this week.
[01:30:56] So we've been doing more regularly what we call kind of our AI use [01:31:00] case spotlights here at SmarterX. So, you know, we hear from listeners, Paul, all the time that one of their favorite things is like when we talk about how we're actually using AI at SmarterX. So we're every week we're gonna try to do.
[01:31:11] Giving you a quick look under the hood at whatever we're working on this week, you know, real use cases that we're either exploring, building, or actually deploying. So Paul, I know you had mentioned you might've had one and I've got a couple I can just really quickly touch on too.
[01:31:25] Paul Roetzer: Yeah, I, I, the one I can give again, I was on vacation for the last, you know, 10 days.
[01:31:29] But like I said, I was in like the back of a tour van and, I had a lot of time while my family was sleeping and I was just thinking, which again, like one, go to Scotland. It's beautiful. two take trips with your family whenever you possibly can. Mike, I know you have a, a young child. if, if I have any parents with young kids listening, like the, my kids are 13 and 14.
[01:31:52] there has never been a single trip I've regretted like, go places with your family, like create memories, create experiences. It's amazing. So [01:32:00] Scott was incredible time with my family. Amazing. my mind was freed. Like it was the first time I stepped away in a little while and just like, just didn't really work.
[01:32:10] Um. Yeah, but it also then gets that inspiration going. So there was one extended trip on the tour where I had this assessment I wanted to build, and I've talked a little bit about one of the assessments I was recently building. This is kind of a compliment, complimentary assessment, which I'll share more detail about in the coming months.
[01:32:27] But I literally just pulled up Claude Code sitting in the van, and I was like, all right, let, let's work on the next one in like three hours in the van. And I mean, I built a working first draft of the thing on my phone in the app. So it's nothing like earth shattering, but the fact I could do this with my phone and an app and go through the entire thing, build a working model, export it into a dock, I could actually edit.
[01:32:55] it, it just easy 30 plus hours of [01:33:00] work traditionally done in three hours in the car. Got back to the hotel that night and I just sat down with my laptop and remember, went to bed and just. Later on it was peaceful and inspiring and not, I didn't feel overwhelmed. I feel like AI psychosis, like it wasn't, I gotta build, it was just, my mind was finally free and I could think clearly and the technology enabled me to do something amazing while I was doing it.
[01:33:24] Mike Kaput: That's so cool. I love that. Yeah, my big one this past week won't be a really a surprise, it's a common one, but definitely just being continually in awe of AI as a deep strategic partner. So I was working a lot of heavy stuff this past week related to helping build out our research agenda at SmarterX, with our director of Research, Taylor.
[01:33:42] And there's a lot of really intense, kind of really reasoning from first principles that these foundational strategy items that are deeply important to getting the direction right on something like this. So the ability to really, in a structured process and methodology sit down and say, okay, I'm gonna gather up all the context [01:34:00] needed, feed this into something like Claude, and then work systematically back and forth with this tool.
[01:34:05] To refine each individual piece of the strategy, but also just the logic behind it. I mean, it's a caliber of thinking I simply could not do on my own, I don't think. So I'd highly recommend Then I took that exact same methodology and said, Hey, this worked really well for this research agenda. I've got three other projects like it.
[01:34:26] Let's apply the exact same methodology to completely different contexts and projects, and also got those done in a fraction of the time it would've normally taken. And more importantly, results I could not do alone.
[01:34:41] Paul Roetzer: Alright, wait, let I show one other one real quick one, Mike. Yeah,
[01:34:43] Mike Kaput: I would
[01:34:43] Paul Roetzer: love that. This is a personal one, so buddy, buddy of mine text me, you know, it's obviously tax time and he was like, dude, I'm getting killed on taxes.
[01:34:53] Like, do, do you have any idea if your, like CPAs talked to you about any tax strategies as a business owner, things like that? And so I [01:35:00] was like, yeah, I'm like fresh off vacation. I haven't slept in two days. And I'm, I'm like, dude, I don't like, I don't really have much here. Like, here's one thing we've tried, like what do you think about this?
[01:35:08] And he goes, yeah, we tried that. And so then I was like, oh shit, you know what? So I go into ChatGPT, and I was like, Hey. And I give it like this basic prompt was I got a friend, he's a business owner trying to figure out tax advantage things. Like what are some, you know, write me a prompt that he could use to do tax planning in his tax bracket as a business owner, whatever.
[01:35:28] It comes back with this amazing prompt. I mean, this is like 1200 words.
[01:35:35] Mike Kaput: Mm-hmm.
[01:35:35] Paul Roetzer: And it's basically, you know, it's like act as a highly experienced US tax strategist, CPA and business advisor who specializes in helping business owners legally reduce their tax burden through proactive planning. Your role is to educate, analyze options, surface questions, strategies, and planning opportunities, follow tax rules.
[01:35:53] And it goes, and then it breaks it down, like boom, boom, boom, boom, boom. Like, go through all these steps. and then it [01:36:00] ends with, and gimme a list of questions I can then send my actual CPA. So I sent this to my buddy and he's like, he used it. He goes, yeah, unfortunately my CPA's really good. And they're like, they've done all these things, but like just that again, it's like that sometimes it's the personal use.
[01:36:14] And then, oh, one other one I was designing a pavilion for our backyard and I used Google Gemini.
[01:36:18] Mike Kaput: Oh, nice
[01:36:19] Paul Roetzer: to do it. And it was amazing. Like it crushed it. So yeah, just some fun personal ones too.
[01:36:24] Mike Kaput: I love that.
[01:36:25] Mike Kaput: Alright, so Paul, we're also doing a kind of regular weekly segment related to spotlighting our courses in AI academy.
[01:36:32] So, if you want, I can kind of tee this up for what we're gonna talk about this week that we've got, available to AI Academy members, if that works for you.
[01:36:42] Paul Roetzer: Yeah, go for it. We did the customer success one I think is what
[01:36:44] Mike Kaput: we're talking about. We did, yeah. So we've had live in AI Academy AI for customer success.
[01:36:48] So this is as a reminder, like a four course certificate series that's built specifically for customer success professionals. And the whole point of this segment is. Just spotlighting the course and kind of giving you something [01:37:00] valuable from it, whether or not you ever take it just as a way to kind of share more of the love that we're, you know, putting together in AI Academy.
[01:37:06] So when I was building this course, kinda what really jumped out to me is I kind of think of it in two ways for these segments. Like why AI matters for this function or industry or segment specifically, and then like how to start operationalizing that. So first step, what really jumped out to me in customer success is just this core, systemic challenge of scale.
[01:37:29] Like if you have, are building a CS team fall? I know you're deep into this right now. Yeah. The only way you historically scale up a CS team is by either hiring more people or piling more accounts onto each customer success manager. So those lead to some really thorny trade offs, right? Like people can get burnt out.
[01:37:48] Your engagement quality from CSMs goes down. top accounts might still get some white glove treatment as you scale. The rest, often do not. Now, what's really cool though is that AI is [01:38:00] starting to break this math in a good way. So instead of scaling by having to raise headcount a ton, or stretching people thinner, you can use AI in a number of different ways to actually scale up the effectiveness of CS professionals.
[01:38:15] Now, again, not getting rid of people or automating them away, but just this ability where you can actually scale without this linear rise in cost. And it's kind of helpful because then CS professionals themselves not only do better work, have better lives and worklife balance as a result, but they can increasingly get out of this like reactive work and start doing much more strategic, proactive stuff that moves the needle and also turns CS more into a revenue center versus like a cost center, which has historically been a big problem.
[01:38:46] So lots of, you know, data points, information and research in this course related to that. But that really is such a core challenge. And what's really cool is, you know, we also teach. Some steps about how to start operationalizing this insight, right? So we actually walk [01:39:00] people through, and I think it's useful even if you just do this on your own, that you start with a really low hanging fruit.
[01:39:06] Like, trust me, the amount of use cases we've got in this course around customer success, there's so much low hanging fruit where AI can no joke, be saving you dozens of hours a a month or even maybe a week at some point. So we really do start from kind of the bottom up and say, look, let's look at where your CS team is spending time on reactive work.
[01:39:26] So these are things like check-ins, qbr, manual scoring, and if you really get smart about making those your first AI targets, you are going to free up people to do so much more time. More time and energy devoted to customers, which is amazing, but also to bigger ticket kind of AI projects and pilots. So the course goes into how to do all that, but even if you don't take the course, I would say literally pull out your calendar tomorrow, screenshot it and drop it into.
[01:39:54] Something like chat GBT or Claude and start talking it through where you're trying to save [01:40:00] time. Then, you know, put a constraint on it. Say, I am, I want to save an hour next week, minimum by next week, let's figure out how to do it. So that's kind of just one little thing that I learned building this out and that we've found customer success professionals who've taken the course so far have also found valuable
[01:40:17] Paul Roetzer: as someone who's trying to architect an AI native customer success team.
[01:40:22] This is highly relevant for me.
[01:40:23] Mike Kaput: Right, right. All right, Paul. So as we wrap up here, we have our kind of regular AI product and funding updates. I'm just gonna run through a bunch. There's obviously a ton since we've been off for a couple weeks, so I'll, I'll hustle through these.
[01:40:36] Paul Roetzer: Yeah. And again, real quick, like, Mike's gonna move fast through these, but honestly, like five or six of these in a regular week, would've been certainly rapid fire topics, if not main topics.
[01:40:49] Like there's a
[01:40:49] Mike Kaput: hundred percent.
[01:40:50] Paul Roetzer: Just because it's a rapid fire at the end here, don't, doesn't mean that some of these aren't very important, that we don't understand the bigger significance, but can only cover so much in a weekly show, I [01:41:00] suppose.
[01:41:00] Mike Kaput: Yes, indeed. So first up, anthropics, annualized revenue has crossed $30 billion.
[01:41:06] That's up from 9 billion at the end of 2025. the number of business customers spending a million dollars or more annually has doubled to over a thousand in under two months. So we've kind of talked about the number, the re, the ways in which that growth is creating some issues for them in this episode.
[01:41:23] Axios is reporting Anthropics usage limits are outpacing OpenAIs and Wall Street Journal reports. They are in talks to invest $200 million in a new private equity venture. Next up, Sycamore is a startup building what it calls the trusted agent operating system for the enterprise, and they just raised a $65 million seed round.
[01:41:43] They are focused on providing infrastructure for deploying and managing AI agents in enterprise settings with built-in trust and compliance. Again, as the topics we just went through in this episode, that should be no surprise. There's a big need for that. Next up, Google released a Gemma 4, which it calls its most [01:42:00] capable open model to date, and it's built from the same kind of infrastructure and research as Gemini.
[01:42:05] And this is really notable because you can actually run this model for free locally. It is extremely capable and super, super powerful. At the same time, meta has introduced what they call Muse Spark. This is the first model from their rebuilt. meta Super Intelligence Lab, Alexandr Wang who runs that noted that the team rebuilt their AI stack from scratch nine months ago and started work on this model.
[01:42:31] so far it's getting decent reviews, it sounds like, in at least certain areas. after kind of some flops from the previous malama model releases, Anthropic Anthropic has launched Claude Managed Agents. These are frameworks for getting AI agent applications to production faster. So the engineering blog post about this details the architecture for essentially decoupling the brain from the hands is how this architecture works, separating the reasoning model from the tools it uses so it can make a agent deployment much more [01:43:00] scalable and reliable.
[01:43:02] Google has added notebooks to Gemini. So these are basically kind of that second brain topic we're discussing. The feature lets users organize project sources and AI conversations into persistent workspaces. Microsoft has introduced multimodal intelligence in co-pilot's researcher feature, which allows it to pull from multiple AI models for deeper research tasks within Microsoft 365.
[01:43:25] So if you have access to this in your account, you may want to check that out. Microsoft and Publicis Group, one of the world's largest advertising holding companies, have expanded their strategic partnership to power the future of AGI Agentic marketing for businesses worldwide. Pika, the AI video generation startup released the beta of its first product with face and voice capabilities enabling AI generated characters that can speak with realistic lip sync and expressions in conversations.
[01:43:55] We alluded to this la this next one. in previous topics, Anthropic published new [01:44:00] research on how emotion concepts function inside CLOs. So it investigates whether LLMs that sometimes appear to express emotions, actually have internal representations that correspond to those expressions. In other news, SpaceX has filed confidentially for an IPO targeting a valuation of more than $2 trillion.
[01:44:20] Google Research published a paper on how to responsibly disclose quantum computing vulnerabilities, which could vary prominently affect cryptocurrency security. So they are making much more of an effort, and I'm sure we'll keep talking about this in the future, about the eventual impact of quantum computing on the current standards of encryption.
[01:44:40] And finally, openAI's introduced a child safety blueprint, a set of guidelines and tools designed to help developers building on openAI's APIs implement safeguards against child exploitation and harmful content involving minors. Okay, Paul, that was a very packed week. One final announcement here. Go to take our [01:45:00] AI pulse survey this week that we had mentioned at SmarterX.ai/pulse
[01:45:04] We're gonna ask this week about some of the big prominent leaks we've had about Claude Mythos, kind of how you're feeling about AI companies operational security. We're also going to ask where you stand on job displacement, and some economists are changing their minds. So, Paul, thanks again for breaking everything down for us.
[01:45:23] This is a packed couple weeks.
[01:45:25] Paul Roetzer: You get under 90 minutes, but I think we slipped over a little bit, but hopefully we stuck with us and it was all interesting. It's, it's just so much to do every week. It's crazy.
[01:45:34] Mike Kaput: Truly,
[01:45:34] Paul Roetzer: but good to be back. next week we're thinking we're gonna have a regular episode.
[01:45:38] I have to travel, but we think we found a way to make it work. So, unless something changes, we'll be back next week with a regular weekly, and then we'll probably actually have a second one next week for Intro to ai. So. yeah, back on schedule, hopefully. All right. Thanks, Mike.
[01:45:57] Mike Kaput: Thanks Paul.
[01:45:59] Paul Roetzer: Thanks for listening to the [01:46:00] Artificial Intelligence show.
[01:46:01] Visit SmarterX.AI to continue on your AI learning journey and join more than 100,000 professionals and business leaders who have subscribed to our weekly newsletters, downloaded AI blueprints, attended virtual and in-person events, taken online AI courses, and earned professional certificates from our AI Academy and engaged in the SmarterX slack community.
[01:46:23] Until next time, stay curious and explore ai.