63 Min Read

[The AI Show Episode 252]: Jensen Huang Resigns from Nvidia, Anthropic’s Mega IPO, AI 2027’s New Warning & HubSpot’s AI Reboot

Featured Image

Jensen Huang is out at Nvidia, and the export-law questions behind his exit could matter more than the affairs that started it.

Paul Roetzer and Mike Kaput unpack why export-law questions around Megaspeed and Singapore chip sales matter more to Nvidia than the scandal. They also cover what Altimeter's push to reinstate Huang signals, and why Nvidia's earnings call has become a read on the entire AI economy.

Then they turn to Anthropic's IPO roadshow and a founders' letter that warns investors the company may act against profit maximization. They also look at the AI 2027 authors pulling their superhuman AI researcher forecast up to 2027. In rapid fire: Fable 5.2, Gemini 4 Carbon and Hydrogen, Brian Halligan's return to "refound" HubSpot, and United Airlines blocking AI agents.

Listen or watch below—and see below for show notes and the transcript.

This Week's AI Pulse

Each week on The Artificial Intelligence Show with Roetzer and 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.

Listen Now

 

Watch the Video

Timestamps

00:00:00 — Intro

00:05:48 — Jensen Huang Resigns from NVIDIA

00:24:06 — Anthropic Launches IPO

00:36:34 — AI 2027 Authors Accelerate Forecast for “Superhuman” AI

00:49:58 — Fable 5.2, Mythos 5.2, Gemini 4 Carbon, and Gemini 4 Hydrogen

00:58:28 — Brian Halligan Returns to “Refound” HubSpot for the AI Era

01:07:03 — SaaStr Sets New Revenue-per-Employee Baseline

01:10:46 — United Airlines Moves to Block AI Agents

01:14:19 — GPTs > Plugins Update

01:16:01 — AI Use Case Spotlight

01:20:08 — AI Product and Funding Updates

 

This episode is brought to you by AI Academy Day 2026.

Join us for AI Academy Day on Thursday, November 19 at 12pm ET / 9am PT. This free event covers how we think about AI learning and development inside organizations, plus what’s next for AI Academy. Register free at SmarterX.ai.


To mark our 250th episode, get $250 off a new AI Academy by SmarterX individual license through November 30 with code POD250. That’s 25% off your first year of the full Academy experience: courses, professional certificates, live sessions, and community.

Leading AI literacy for your team? Buy 25 or more licenses by November 30 and Paul Roetzer will run a private one-hour virtual session with your team. Buy 250 or more and Paul will come to you for a private in-person session in the first half of 2027 (US only). Existing business members adding 25 or more licenses qualify too. Join at academy.smarterx.ai, or contact our team at support@smarterx.ai about business licenses.

Read the Transcription

Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.

[00:00:00] Paul Roetzer: We're just increasingly going to be living in this world where the models are getting way smarter, their capabilities are expanding exponentially. There's going to be this huge impact on people and companies, but not everybody's gonna experience it in the same way, and some people are gonna be oblivious to what's going on.

[00:00:19] 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.

[00:00:39] 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 2 52 of the Artificial Intelligence [00:01:00] Show. I'm your host, Paul Roetzer, along with my co-host Mike Kaput, we are recording on Monday, November 16th. the Jensen Huang situation has progressed. by the time you listen to this, there might be even more interesting things happening, but we are definitely gonna revisit what's going on with Nvidia.

[00:01:19] it is, it was a wild weekend, just honestly in the news. So we'll give you the latest as of Monday morning and then, yeah, Tuesday when this is dropping as their earnings report day. So it's just, it's gonna be a very interesting week in the world of ai. So we'll get into all of that. Also, this week is AI Academy Day.

[00:01:39] So this is happening on Thursday, November 19th at noon Eastern Time. This is our, upcoming live event about the next phase of AI education. So, as AI models and agents get more capable, the way we build AI skills needs to keep evolving too. And that's the idea behind what we're building at AI Academy by [00:02:00] SmarterX.

[00:02:00] So again, November 19th at noon, Mike, and I'll join our head of learning. Jessica Miller to preview what's next. We're gonna talk about how, how, learning that adapts to your role, goals and progress and share new, new courses and live learning experiences Coming to AI Academy. I'm actually like, so I'm traveling the next couple days and I am, uh.

[00:02:20] Finalizing some decisions around the things we're gonna be announcing. So this is gonna be like, this is gonna be real time stuff, but we, I sort of previewed it in the essay I wrote last month. but basically like the big focus moving forward is dynamic learning and applied learning. And we're gonna try and take both of those to the next level in terms of the value we can create for organizations and individuals on an ongoing basis with our academy.

[00:02:44] So I'm extremely excited about the direction it's going and, yeah, so you can join us even if you're not an AI Academy member or even don't plan to be, it still could be a really valuable event for you. We're just gonna talk about, transformation within organizations and what [00:03:00] that looks like moving forward, and then how we're trying to solve it, basically.

[00:03:04] So yeah, even if you're not a member, I'd suggest get, you know, join if you can. It's just SmarterX dot ai slash academy day and you can register for AI Academy. It is a free live event. They'll be an ask us anything at the end. and again, we're gonna kind of go through what we see as like the future of education based on all the conversations we've had with our customers, just people in the industry and, yeah, so exciting stuff.

[00:03:32] I'm not a hundred percent sure all the things we're gonna announce yet. but I'm, I'm zeroing in on them and so hopefully over the next 48 to 72 hours I sort of lock in on all the final announcements.

[00:03:42] Mike Kaput: That's what makes it exciting.

[00:03:44] Paul Roetzer: Well, yeah. Says the guy who's presenting with me who has no idea what we're presenting yet.

[00:03:49] Mike Kaput: Yeah, yeah. Stressful, exciting. You know that there's a difference. I don't know, maybe.

[00:03:54] Paul Roetzer: Well, that's you. I think you saw it, Mike, like the, the AI score for business leaders presentation I did last Friday, where if you didn't join us, we'll put a link in. It's on demand now on YouTube. We put the whole thing on YouTube.

[00:04:07] I literally finished that presentation 30 minutes before we started, and it's because there was things I just wasn't. Sure yet, like how I wanted to present them. I have like nightmares about that where you're like finishing a presentation right before you go on. But it, I don't know. It worked out. I was, you know, I think it was a really good event and I'm excited about the things we're building and sometimes you just, there needs to be a forcing function to make the final decisions about the things you've been thinking about for like 12 months.

[00:04:34] So yeah, Thursday is my forcing function. We will, we will be going one way or the other. So SmarterX AI slash Academy Day and you can join us for that on Thursday. Alright, Mike, we got the pulse survey. So this is, we have the results from this survey. Has AI expanded your role into work that typically belongs to adjacent roles?

[00:04:55] So this is the idea of like, if you're in marketing, all of a sudden you're able to do some sales stuff. [00:05:00] So 44% said yes, my role now blends significantly with horizontal and adjacent or adjacent roles. 33% said somewhat. I occasionally take on adjacent work. 21% said no. And then there was apparently a sliver.

[00:05:15] People said, I don't use AI for work.

[00:05:17] Mike Kaput: Okay.

[00:05:18] Paul Roetzer: I wonder how they ended up in that survey.

[00:05:20] Mike Kaput: Yeah, no kidding. Yeah. And all the people we always ask too, all the people, our podcast listeners. So we've got someone listening that's not using AI for work and that's okay.

[00:05:28] Paul Roetzer: Maybe they're just getting started. That's cool.

[00:05:29] Mike Kaput: Right. Damn.

[00:05:30] Paul Roetzer: Alright, so, so yeah, SmarterX AI slash pulse. Michael, remind us at the end the new survey will be up and you can take part in that takes literally about 10 seconds to take the survey. Alright, Mike, let's, let's dive into the Nvidia saga, which just keeps sort of expanding.

[00:05:48] Jensen Huang Resigns from NVIDIA

[00:05:48] Mike Kaput: Yeah, Paul, so last week we had covered.

[00:05:51] NVIDIA's investigation into national security concerns surrounding Jensen Huang. we had talked about his, [00:06:00] infidelity with Alice Huang, who is suspected to be a Chinese spy, but it, since that episode. Aired. The Chipmaker has announced that Huang has resigned as CEO after 33 years leading the company.

[00:06:14] So for now, chief Financial Officer Colette Kress, is serving as interim CEO, Nvidia, searching for a permanent replacement. Huang had a acknowledged infidelity after reports of that relationship with Alice Huang. She was a former executive at Megaspeed, an Nvidia customer in Singapore Reuters, and the New York Times subsequently reported that Huang admitted to additional relationships with Nvidia employees after this one broke.

[00:06:40] An open question is the role of Huang Alice Huang. We have to differentiate here. Ms. Huang. Ms. Huang, yes. who some reports do allege WA had ties to Chinese intelligence, which is how whole kind of drama kicked off not just the infidelity, but national security concerns. However, [00:07:00] on this past Thursday, attorney General Todd Blanche said.

[00:07:03] That the Justice Department had not concluded whether Ms. Huang was a Chinese spy. China denies the allegation. Megaspeed says it operates lawfully and has hired outside lawyers to investigate the whole thing. However, US authorities are investigating Megaspeed over suspected efforts to get restricted chips to China.

[00:07:22] Three house committee chairs have requested a closed door interview with Jensen on December 1st and threatening a subpoena if he does not cooperate. Meanwhile to significant Nvidia shareholders, Brad Gerstner's, Altimeter Capital Management, and Gavin Baker's Atreides Management, sent a joint letter to NVIDIA's board this past week.

[00:07:44] They urged directors to take their time choosing NVIDIA's next CEO and keep Huang's possible return on the table. While the facts surrounding his departure become clearer. Elon Musk has separately said he's signing a petition to bring Huang back. I think Paul [00:08:00] recently, as of yesterday, there was even more talks and like higher odds on possibly people trying to rally to get Jensen back.

[00:08:08] Like, there's a ton that's happened since the first segment where what are you paying attention to? What are you focused on here? Can you make sense of this mess for us?

[00:08:21] Paul Roetzer: Yeah. So I think I mentioned on the last episode that I assume someone in the government, obviously wanted this information out in a very coordinated way.

[00:08:28] it appears that it may have come from Marco Rubio, the Secretary of State. Like that's a lot of the discussion is that he tried to. Get the administration, take this stuff seriously, and they, they didn't. And and so it seems like, you know, maybe there's a coordinated effort to do this. So I'll, I'll try and break this down of like why we all should care, because, you know, obviously, you know, at a high level it's you know, tech executive having affairs, like you may or may not care about that.

[00:08:59] Mike Kaput: Right.

[00:08:59] Paul Roetzer: [00:09:00] but I'll sort of break down the other implications of what's going on. So there's a lot of articles, a lot of things you could go read on this. There was one that I spent some time with that I thought did a nice job. This came out on, I think this was on Sunday, November 15th in the New York Times.

[00:09:16] And you know, when they have like six journalists in the byline that it's going deep because Oh yeah. Basically. Oh

[00:09:22] Mike Kaput: yeah.

[00:09:22] Paul Roetzer: All tapping all of their sources to find out what's going on.

[00:09:25] Mike Kaput: Yeah.

[00:09:25] Paul Roetzer: So this article was, anyway, it's actually SA Saturdays, November 14th. Jensen Huang carried a secret for more than a year.

[00:09:32] It led to his downfall is the title. So I had kind of gone through and. Created my own outline of like the things I wanted to touch on here. And it just so happens this article touches on numbers. So I'm gonna cite this article a couple times and so I wanna like lead off with that. So the, what are the major issues here?

[00:09:49] So the first is the affairs, concluding with a potential spy, which he denied for months. And then finally admitted to general counsel at NVIDIA that he had in fact [00:10:00] had this affair. And then, after that it was the coworkers relationships that came to light. And that is appears to be one of the things that triggered the board to finally take action as he lied about those as well.

[00:10:12] the lying about the affairs is probab potentially at least a bigger deal than the affairs themselves. So in the New York Times article, it said the revelation of another infidelity showed that Mr. Huang, who had extolled the company's code of conduct rules over the years had been violating them. First there was the relationship with a customer, then there was the relationship with an employee.

[00:10:35] But in the world of corporate governance, the worst thing a chief executive can do is lie, especially in a crisis. It is the reason, they get fired. So that was again from the New York Times. Then there's the bigger issue here. So again, like let's say you can look past the affairs and you can look past the lying about the affairs for months or maybe over a year.

[00:10:56] There are the concerns around bypassing export laws, which is [00:11:00] a much likely larger issue here. Assuming he wasn't knowingly or unknowingly passing intelligence to the Chinese. Through the lady he was having the affair with. Okay. That's a pretty big deal too. But the concerns around bypassing export laws, so again, I'll go back to the New York Times, companies like Megaspeed were legally buying American ships and building AI data centers.

[00:11:22] that rented computing power to Chinese companies. But the US government suspected that some including Megaspeed, were breaking the law by funneling chips to China. NVIDIA's sales from Singapore were swelling. This is going back to fall of 2025 when the New York Times was first investigating this and wrote about this.

[00:11:42] It said, by last summer, the Island city state accounted for 22%. Of the company's quarterly total sales, or $10 billion, making it the second largest chip buying region after the United States. So Singapore, $10 billion in a single [00:12:00] quarter after the article about Megaspeed and Mr. Huang's relationship with Alice Huang appeared in October, 2025.

[00:12:08] A year ago, Nvidia changed how it reported geographic sales. Singapore vanished from its financial disclosure disclosures as it started reporting sales based on customers headquarters rather than billing location. The company had previously said that large customers billed through Singapore where taxes can be lower, but that products are always, almost always shipped elsewhere.

[00:12:32] So. That seems to be the bigger issue here is like there's this ongoing investigation and this is what Rubio was like pushing on with the administration is like, something's not right here. Basically it seems to be what was going on. So then what are the implications? So obviously, Nvidia stock is not doing well and a lot of people's retirement accounts are based on Nvidia stock.

[00:12:54] a lot of portfolios of, investment funds are based [00:13:00] on this, like it's a very important stock to the economy and to individual investments at that broader economy level. The New York Times article said Nvidia, which was in many ways propping up the AI boom, which in turn was propping up. The US economy is operating without a man.

[00:13:16] Many considered to be one of the most indispensable executives in the history of American business. This is why the Silicon Valley Brotherhood is sort of rallying around him, is they're basically saying. Regardless of what he did and how many affairs he had, it doesn't matter because we need him is in the US economy is is pretty much their argument, is kind of how it comes across.

[00:13:38] another thing is this idea of propping up the AI boom. Nvidia was basically bankrolling a lot of these neo cloud companies. And a lot of the labs themselves, like basically. Bank rolling, like, OpenAI, build out of data centers and stuff like that. So they were like backstopping where the banks wouldn't step in and fund.

[00:13:58] Jensen was, and [00:14:00] it sounds like in some cases it was like through handshake deals, like Yeah, it's just, it's like this cascading effect. so the other thing is, we've talked about this and I've expressed like uncertainty around why Jensen was taking the positions He was with open source and chip exports.

[00:14:17] Like we talked I members specifically like a month or two ago where he was on, was it Dwarkesh? I think where they got like, I think so, yeah. They kind of got in an argument about it.

[00:14:25] Mike Kaput: Yeah.

[00:14:26] Paul Roetzer: and Jensen was feeling was like, everything's fine. Like it's whatever. Like we're, everybody's gonna have the technology anyway.

[00:14:32] Let's just open source. And I was like, it's just a weird position. So he is like the major advocate for open source and chip exports, which again goes to the Silicon Valley brotherhood of like many of them want. The open source movement pushed and Jensen was the guy with TR to in Trump's ear that was like getting that.

[00:14:53] Which leads to the very influential nature Jensen had with Trump and the infighting with administration [00:15:00] officials. So the New York Times again, it said it has brought into public view a few among several senior administration officials and Commerce Secretary Howard Lutnick on Friday, treasury Secretary Scott Bessent told CNBC that his department had been pushing commerce for months to crack down on the flow of US technology to China.

[00:15:20] So there's like, now people are getting thrown under the bus. It's like, Hey, it wasn't us. We were the ones telling Trump and telling these other guys to like. Stop what was happening and they just weren't letting it happen. And then changes to export rules could happen fast, like it's bipartisan so that the New York Times article said administration officials are now considering, reconsidering AI export regulations, which could hurt some Silicon Valley businesses, including the investors who are pushing for Jensen to be back in.

[00:15:47] But, tilt AI race with China decidedly in America's favor. one interesting note here, I kept asking myself like, who benefits from this? Not like they had any role in this happening, but like, who would be the beneficiary [00:16:00] of this? Google is the one I just kept coming back to with their tpu. So like, who has an alternative here?

[00:16:05] If Nvidia stumbles, it seems like Google. Would be the most logical beneficiary, but I haven't really thought super deeply about that one. but yeah. earnings call Tuesday night, you know, appointment viewing, like I, I, yeah, they already said they're probably not gonna provide guidance because of the uncertainty, but the interim CEO will will be on.

[00:16:28] you mentioned this push to have him reinstated by Gerstner and Baker two guys we've talked a lot about on the show, the letter from Altimeter said, given the extraordinary success of Jensen Huang's leadership over a long period of time, and given the massive uncertainty associated with recent events surrounding his departure, meaning maybe she wasn't a spy, is kind of the thing that they're mainly alluding to.

[00:16:50] Yeah. We believe it is in the best interest of all stakeholders and certainly the shareholders of Nvidia for the board to pursue a slow measured and deliberate process regarding the future [00:17:00] leadership of Nvidia. That leaves all options on the table, including the potential return of Jensen Huang.

[00:17:05] Apparently Jensen spent the weekend like holed up with these guys. Contemplating how to get him reinstated. Like this is actively a thing. Yeah. That's why by Tuesday, who knows? And then it said there are more questions about recent events than answers. We believe the board, the board should avoid a brush to replace Jensen that undermines and thwarts the best interest of shareholders.

[00:17:26] Stephanie Ruhle, a journalist put this on X and I thought this was like, again, sort of a come back to reality moment. She said the Wall Street Journal reports that Megaspeed, the company led by Jensen Huang's, former mistress or maybe current mistress that we don't, we don't even know that Alice Huang lined up nearly 12 billion in top tier NVIDIA chips for Chinese buyers between March, 2024 and June, 2025.

[00:17:49] If those chips actually reached China without licenses, that would be a serious breach of US export law. So like we're not just talking about affairs here. Like [00:18:00] if, if. If the Silicon Valley Brotherhood is like, yeah, whatever. He slept with three different women while he was married. Like, and one of 'em was his coworker and one of 'em, I think two were coworkers and one was Yeah, a like who cares is their basic positioning.

[00:18:11] And that's fine. Like if that's their belief, that's not why he's fired though. Like it, it, I don't know. So what happens next? The FBI has an o obviously has an investigation going on that they've had for like a year. This is involving Secretary of State, Marco Rubio, who personally raised concerns to the president about the surveillance information they have, which the president appears to have dismissed.

[00:18:35] We were correct in the whole like, member last episode, Mike. I said it was odd that like he wasn't on that flight and then all of a sudden he was Right. Okay. So Rubio is the one that didn't want him on that flight. And when Trump saw the news, this came out in the, I'm not making this up, this came out in the news that.

[00:18:52] When Trump saw the backlash as they were on their way to China He called and had Jensen meet them in [00:19:00] Alaska. So like Trump went around the State Department, the Department of Justice, the FBI, and personally put him on the flight, even though all this was going on, I believe Jensen already agreed to appear before Congress December 1st is I saw that now that could change.

[00:19:15] I, things can obviously move, is a fluid situation. I can't imagine they reinstate him before this investigation plays out because it sounds like there's some serious things going on here That Right. That NVIDIA's board could be liable for. Like, it just feels like this is gonna play out for a while.

[00:19:37] I can't imagine they reinstate him though while they're, this is playing out le leaving it as an option. Fine. Like maybe you decide to keep an interim CEO for a while and let this like play out and he functions as an advisor or something. But putting him back as CEO, like the blip at OpenAI where like altman's back in after 72 hours, I don't think it's that.

[00:19:58] Mike Kaput: Okay. So [00:20:00] just a couple threads to kind of wrap up here. So first it sounds like there's obviously always the possibility until confirmed that the real danger is like, okay, Jensen intentionally or actively shared secrets with Chinese intelligence. It seems nothing has come out about that. It seems like if that was suspected, Jensen might be in.

[00:20:21] Prison keeping in jail. Yeah. So, so that's the, put that aside. But the bigger issue is it sounds like from these numbers, there's a real chance that for whatever reason in the pursuit of revenue, possibly a lot of sales to Singapore were being routed around US export laws because those chips then were going to China, whether that was facilitated by Chinese intelligence or not likely so.

[00:20:45] But regardless, that is a huge deal and what they are likely investigating.

[00:20:50] Paul Roetzer: Yeah. So it sounds like the US government is fairly confident that they have a case that chips were illegally ending up in China through companies like [00:21:00] Megaspeed. And now the question is, did Jensen or Nvidia know that that was happening?

[00:21:06] And if that. If they determined that they did and they can prove that they did, there are much bigger problems for Nvidia than their CEO resigning or being asked to leave, like,

[00:21:17] Mike Kaput: right.

[00:21:18] Paul Roetzer: And so I think that's the big question now. And that's why I'm saying like, I can't fathom that you could put the CEO back in play if these departments in the government believe that there's something worth investigating here, which they obviously do if they've been at this for a year already.

[00:21:34] Mike Kaput: And with the added wrinkle that while obviously Nvidia can still make chips without Jensen and power, the AI economy so much was wrapped up in jenssen's literal personality and personal relationships to the point of your, you made about handshake deals that, you know, ultimately in a long enough timeline, nobody's irreplaceable for the most part, but like he is pretty [00:22:00] damn close to it sounds like, in terms of probably

[00:22:02] Paul Roetzer: as close as you could get.

[00:22:03] Right.

[00:22:03] Mike Kaput: Okay.

[00:22:04] Paul Roetzer: Gotcha. Yeah, but I mean, it's like. It's Steve Jobs esque with Apple, but like, but a much larger part of the economy. Like it literally is dependent. We've said this for the last couple years, like Nvidia. Their earnings is like every three months. You just wait to see what happens because it's the greatest indicator of is this AI boom going to continue.

[00:22:25] Like, because their earnings are basically projecting out like two to three years of build and it's like, hey, as long as everything's still going and you know, they're still building and the demand's still there, then you know, at least the next couple years we're still gonna be riding this wave. So, I don't know, I don't know what happens Tuesday night.

[00:22:42] but you, you know, if, if you wake up Wednesday morning and your NVIDIA investments aren't looking so good, you're, you'll know why go check the earnings calls and the response to 'em. So, I don't know. The board's gotta do something. They have to have some roadmap in place by Tuesday night. They need some clarity.

[00:22:59] Even if they [00:23:00] don't provide guidance, they've gotta provide some clarity about what's happening, what the internal investigation process looks like. What is the real state of Jenssen's involvement with the company? Is he just on. Is it, is it, is it a resignation and he's done or is there a possibility once this all plays out, they'll bring him back?

[00:23:15] If they say that, I would imagine the stock might, stabilize. Yeah. That if they, if they indicate there's a chance, this is be fine and it'll come back. I think things look better if they say we have concerns and like, this doesn't look good for Jensen. and I keep coming back to that Trump thing on 60 Minutes, Mike, with the

[00:23:34] Mike Kaput: Yeah.

[00:23:34] Paul Roetzer: You know?

[00:23:34] Mike Kaput: Yeah.

[00:23:35] Paul Roetzer: Give your wife a kiss 'cause you're going to jail tomorrow. You're gonna look, I don't, I don't know who else he would be referring to other than this. And I, again, I don't, not to speculate like what Trump was saying, but we were running through the thing like, who else could he be talking about?

[00:23:50] Like the AI leaders going home and kissing your wife and you're going to, the fact that he did that interview and said that the week all this came out, it just,

[00:23:59] Mike Kaput: yeah.

[00:23:59] Paul Roetzer: [00:24:00] It's weird.

[00:24:00] Mike Kaput: Yeah. It seems unlikely it was someone else with that strong a message. Yeah.

[00:24:04] Paul Roetzer: Yes.

[00:24:06] Anthropic Launches IPO

[00:24:06] Mike Kaput: All right, so our next big topic somewhat related here.

[00:24:08] We have been following Anthropics plans to go public. The company is now days away from its stock market debut. So Anthropic expects to set its final IPO price on Wednesday, November 18th. So just a day after you are listening to this. It is set to begin trading on the Nasdaq on Thursday, November 19th, under the ANT.

[00:24:30] This past week, the company launched its investor roadshow. They're offering a 825 million shares within a price range of 70 to $78 a piece. At the top of that range, the public offering would total about $64.4 billion. Most of those shares are being sold by Anthropic, but roughly 24 million shares are sales by existing stockholders.

[00:24:52] There's a plan, $26 billion private sale of stock to seven companies, including Google, Nvidia, and Microsoft, which would bring the combined deals [00:25:00] to about $90.4 billion at the top of the price range. Bloomberg says that combined figure would exceed the amount raised in SpaceX's IPO. This funding comes alongside some enormous spending commitments that we've mentioned in the past.

[00:25:13] Anthropics prospectus lists roughly $901 billion dollars in purchase commitments and leases almost all four. Computing and infrastructure extending through 2045. About $71 billion is due in the year following September 30th, 2026. Some agreements have. Conditions or termination rights. So it's not necessarily they're on the hook for sure for all of that.

[00:25:37] Bloomberg reports, the company is, targeting a valuation of about, $1.4 trillion on a fully diluted basis, which includes potential shares from things like stock options. Investor interest is running well above the available shares. Bloomberg reported that this past week on November 11th, that the IPO was multiple times oversubscribed, meaning prospective investors had requested [00:26:00] several times the amount of stock on offer.

[00:26:02] It's worth noting the company's revenue is also growing rapidly. Anthropic reported an $81 billion annualized revenue run rate. So projecting annualized revenue from current numbers as of October 31st, that's up from $71 billion of annualized revenue run rate, projected a month earlier, and it was $65 billion at the end of July.

[00:26:26] For the third quarter itself, Anthropic reported $17 billion in revenue and an operating loss equal to about 3% of revenue. It's adjusted operating margin, which excludes certain expenses was 15%. So Paul, I'm curious, like thoughts on the Anthropic Roadshow, IPO, how I, the big question for me is how, if at all, does the Nvidia debacle.

[00:26:48] Affect sentiment around this IPO? Is that a concern?

[00:26:52] Paul Roetzer: It doesn't seem to be. I mean, all indications are there. there's enormous demand for the stock [00:27:00] and for the IPO. so yeah, I don't know. We'll see Tuesday night leading into the pricing on, on Wednesday and the offering on Thursday, like where it goes.

[00:27:10] Um. You did a good job just kind of covering the numbers. it's hard to wrap your head around. I mean, literally you're just like, these are just crazy numbers and it's hard to think about them. So I wanna focus on the IPO page. We'll put the link in. It's just ipo dot Anthropic dot com is, is where you can go and kind of check this out for yourself.

[00:27:28] But I specifically spent time really thinking about the founder's letter, Mike, on that page. Yeah. So the IPO page leads with AI will have a vast impact on the world. Anthropic is a public benefit corporation dedicated to securing its benefits and mitigating its risks. So I'm gonna go through just a, a few excerpts from the founders' letter 'cause I think it helps frame.

[00:27:50] What this company is about, what their founders believe and the things that they're focused on. And they're very clear with potential investors the [00:28:00] risks associated with investing in this company and with AI overall, almost to a fault. Like there's, I I've never seen anything quite like it in terms of the risks associated with a company and their IPO.

[00:28:12] Okay, so it starts off with, we believe that AI is one of the most consequential technologies in human history. If managed well, it could rapidly accelerate scientific innovation, greatly increase economic growth rates, bring cures for major diseases within reach and solve problems that have resisted generations of human effort.

[00:28:32] There are four things that we would like any investor in Anthropic to understand. One, the incredible potential of ai. Two, the risks as the risks We much must address three Anthropics unique approach to building a company and four operating in the world, and the ways in which our behavior may differ from that of a purely profit maximizing company.

[00:28:54] So I'm just gonna highlight a, a, a piece of each of those four things, Mike. So an AI's [00:29:00] potential section and the whole thing's like eight pages. Yeah. You can read this thing in like 10 minutes. it says it is anthropics thesis that the exponential in cognition is likely to continue eventually reaching the level.

[00:29:12] What we have called a country of geniuses in a data center, which are millions of virtual agents with the ability to revolutionize every area of human endeavor software, physics, math, biology, medicine, finance, economics, commerce, robotics, military. And intelligence capabilities and much more. Anthropic has always focused first and foremost on working with other businesses because we understand that the extraordinary benefits of AI will mostly derive from the way our models are deployed by enterprises and other organizations across the world.

[00:29:45] Okay, AI risks number two. To build AI is to harness cognition itself and anything so powerful can be used for ill as well as good. Anthropic is a deeply values driven organization, which is what they often get criticized for [00:30:00] actually by outsiders. founded in large part to help the world realize AI's benefits and avoid its dangers.

[00:30:06] There are four broad categories of risks, misuse, loss of control, authoritarian abuse, and economic disruption. I'm not gonna go into each of those, but in the founder's letter you get kind of the high level of what their concerns are with each of those. That's worth reading, especially the economic disruption One I thought was interesting.

[00:30:26] Okay. Anthropics unique approach. As founders and business leaders, we believe deeply in the power of entrepreneurial capitalism to deliver massive benefits to humanity. But we also believe that for a technology as consequential as ai. Entrepreneurial capitalism must be balanced and augmented by mechanisms that protect the common good and share the up upside, upside as widely as possible.

[00:30:50] Everything about the way we have built Anthropic from our culture to our corporate structure has been designed to ensure we can act in the public interest when the situation demands it. [00:31:00] Our theory of change has always been what we call race to the top. We've talked about that many times on the podcast.

[00:31:05] That is like their very firm belief is like they gotta get there first. They said by showing that it is possible to succeed commercially while acting in public interests. We hope to persuade other companies to do the same. We will make mistakes and we expect to learn from others as much as they learn from us.

[00:31:23] If our competitors adopt our practices and improve on them, we consider that a success. one example there might be like mechan, mechanistic, interpretability, where they share a lot about their studies of like looking into the models and trying to understand them, and then other people kind of follow along.

[00:31:38] Alright? Then they went into investor expectations. Although we believe that AI as a field and Anthropic in particular represents one of the greatest economic opportunities in history, we also believe that as important to state clearly the ways in which our decision making may run counter to investors expectations.

[00:31:57] These fall into two categories. Business [00:32:00] bets that derive from AI's wild exponential trajectory, but seem strange or unorthodox in isolation and decisions that depart from profit ma maximization in order to appropriately balance the public good in the first category. Anthropic will often make forward investments, especially in compute, like building out data centers and buying Nvidia chips and things like that that may seem irresponsibly large at first glance.

[00:32:25] This derives from our belief based on the scaling laws, which their founders wrote the paper for. That AI's economic value is on an exponential curve and that the expected near future economic value of our products will therefore be much larger than the current value. Now keep in mind on this one, Mike, we talked, I think it was on the last episode, um.

[00:32:46] About how when they're making these bets on compute, how many chips they need to buy, you're basically making forecast two to three years out. Yeah. And every one of them in the last like three months has said they wish they would've bought more two years ago. Yeah. [00:33:00] So that's what he's saying. He's like, we're gonna make investments where you're gonna be like, that's crazy.

[00:33:04] And then two years is gonna arrive and you'll be like. Damn it, we didn't buy enough. in the second category departure from profit maximization, we are committed to standing up for our values at critical points, even when it is costly for us commercially. The example they gave here was Claude Mythos preview, which happened in, I think that was like what, February or March, Mike, when all that started going out with the government, yeah.

[00:33:26] They said it was the most capable model we had built at the time. We judged that its cap, its abilities, particularly in cybersecurity, had advanced to a point where releasing it widely without new safeguards would've carried significant risk. We therefore chose not to make it generally available and offered its capabilities to the public as Claude Fable five, only after we had built additional safeguards, the delay meant foregoing substantial revenue and seeding ground to competitors.

[00:33:52] We also gave away around a hundred million in Claude credits to cyber defenders. But we believe it was the right decision and we expect to face similar [00:34:00] decisions again as our models grow more capable. And then they also said, and I thought this was an interesting maybe shot at OpenAI, we have also committed to keeping Claude free of advertising.

[00:34:10] parting thoughts more broadly. However, we encourage you as a prospective owner of our stock not to think of your purchase in purely transactional terms. We hope you will see it instead as an opportunity to be part of something greater. The advent of machine intelligence is one of the seminal moments in human history as one of the companies at the very center of that moment, we have a rare opportunity to shape it for the better.

[00:34:35] Our aim is to help humankind chart a safe course through this technological evolution, open new frontiers of discovery and bring forth an age of miracles. Everyone who owns our stock, whether an institution or an individual becomes part of that effort, we hope you'll share our ambition to see it through.

[00:34:52] And this was signed by Dario Amodei, Daniela Amodei, his sister, Tom Brown, Jack Clark, Jared Kaplan Jared [00:35:00] Kaplan, Sam McCandlish, and Chris Olah, who was the one we talked about that, was with the Pope, basically. Yes. That tried, you know, talked about consciousness of the Pope and things like that. So 24607 listed as co-founders that signed off on that letter.

[00:35:16] Mike Kaput: So it sounds like, correct me if I'm wrong, Anthropic is like, Hey, we're playing the game of being a capitalist company, going public, but also we're different.

[00:35:26] Paul Roetzer: Yeah. We may make decisions that you're gonna, as an investor, be like, I don't like that decision. That's not gonna help the stock price. And they're saying, you should know that in advance and you're along for the ride.

[00:35:38] Mike Kaput: That is interesting. I'd be curious to go kind of research how, this is not common to have this kind of no language and decisions in a prospectus for a roadshow. Most of this would, you know, companies seem to have like their mission and what the why they do it. But this is a, seems like quite a bit different than your typical company going public

[00:35:59] Paul Roetzer: for sure.

[00:35:59] Yeah, it [00:36:00] is a very unique position. It's a very unique company. It's, they have very unique beliefs about the future and what these models really are and what they're capable of. I think they're aligned on the capabilities with the other AI labs, but how they think about these models and their role in society, I believe is different.

[00:36:20] in a lot of ways.

[00:36:23] Mike Kaput: And to your point, it doesn't seem like that makes investors nervous because people are beating down the door to try to get into, get some stock here.

[00:36:32] Paul Roetzer: Yeah. Nice.

[00:36:34] AI 2027 Authors Accelerate Forecast for “Superhuman” AI

[00:36:34] Mike Kaput: All right, so third big topic this week. So last week we covered METR, the nonprofit, their assessment that AI is accelerating research within Anthropic.

[00:36:46] But now that report has prompted the researchers behind the AI 2027 report, which we've talked about several times. We'll refresh your memory on that again. It has prompted them to move [00:37:00] their timelines forward. So AI 2027 was this kind of big scenario planning research project published in April, 2025, and it laid out this scenario in which AI systems could automate AI research, create increasingly powerful successors, and eventually we would reach super intelligence by their prediction at the end of 2027.

[00:37:22] So the authors of this scenario are kinda researchers, AI insiders, who are basically trying to forecast and project what could happen if we get. Runaway intelligence explosions here. So they lay out some hypotheticals about what becomes possible in the world if that happens. So they actually, you can go read the full thing where we'll link to it in the show notes we've talked about in the past.

[00:37:44] They theorize, they kind of play out these fictional scenarios. They theorize, for instance, in AI lab running hundreds of thousands of copies of an AI researcher working at about 50 times human thinking speed, and making a [00:38:00] year's worth of algorithmic pro progress every single week Now. If you go read this, it sounds really sci-fi.

[00:38:07] There's like no question about that. But the point here is these authors are trying to game out what a possible future looks like when AI gets good enough to conduct AI research itself. When it's able to then develop more capable versions of AI without necessarily human involvement. And then those systems speed up.

[00:38:28] The next round of research and repeat, rinse, and repeat. So that is recursive self-improvement and it causes a potential intelligence explosion. Now that's the important part because regardless of their forecasting, like that part is something people appear to be concerned with. So this past week, the AI Futures Project forecasters now said they expect a superhuman AI researcher.

[00:38:54] In 2027, assuming there's no deliberate AI slowdown, so that means [00:39:00] AI that can fully automate coding at an AI lab while matching its best human researchers at choosing research directions, designing experiments, interpreting results. So what changed their thinking kind of bring this full circle was that meter assessment of Anthropic.

[00:39:16] So they said, look, if those estimates are right. AI is already accelerating research at a level that the original AI 2027 scenario did not reach until roughly April, 2027. So by that measure, the researchers behind this say that we're actually about five months ahead of the original AI 2027 scenario. So to put this change in perspective, one of the key researchers here, Daniel Kokotajlo, who we've talked about before, had put the, 50% probability for a superhuman AI researcher by the end of December, 2028.

[00:39:52] He now puts it at May, 2027. So basically, the authors are concerned here that a system [00:40:00] pursuing goals and iteration and recursive self-improvement that humans don't necessarily intend or control. Could create huge problems. They actually, when they updated this past week, the AI 2027 forecast, they actually accompanied it with a call for a pause or a substantial slowdown in frontier AI development.

[00:40:21] So Paul, that's all you know. We've got. The mu not mundane, but the very human concerns around IPOs Jensen leaving, and now we've got these researchers saying, Hey, we had this crazy sci-fi scenario and prediction. We're actually too slow.

[00:40:37] Paul Roetzer: Yeah. So I actually went back, Mike and I looked it up. We, so we talked about the AI 27 research on episode 143.

[00:40:44] Yeah. So it was March of 2026. That was March of 2025. That, that would've been 2025, I think it would've been

[00:40:51] Mike Kaput: 25. Yeah. Right after this comes out,

[00:40:53] Paul Roetzer: I think. Yeah. And so at the time we covered Dwarkesh, the podcast had also interviewed the [00:41:00] authors. And so I grabbed a couple of the excerpts that we'd highlighted back in March of 25.

[00:41:05] And so they said on that podcast, we predict that the impact of superhuman AI over the next decade will be enormous exceeding that of the industrial revolution. The CEOs of OpenAI, Google, DeepMind, and Anthropic have all predicted that AGI will arrive within the next five years. Sam Altman has said OpenAI is setting its sight on quote super intelligence in the true sense of the word and the quote, glorious future.

[00:41:30] It's tempting to dismiss this as just hype. They said this would be a grave mistake. It is not just hype. We have no desire to hype AI ourselves. Yet we also think it is strikingly plausible that super intelligence could arrive by the end of the decade if we're on the cusp of superintelligence. They said society is nowhere, near prepared.

[00:41:52] Very few people have even attempted to articulate any plausible path through the development of super intelligence. So that was the [00:42:00] whole point at that time, is they were, they might have been wrong in their early projections, but that were at least trying to say, what would this look like? And so the people who have thought maybe most deeply about this are telling you.

[00:42:13] We think we were conservative. Yeah. And at the time it seemed crazy.

[00:42:18] Mike Kaput: Crazy. Yeah. Totally crazy. Yeah.

[00:42:20] Paul Roetzer: So I think the part I took away is that like we're going to be increasingly uneven, unevenly distributed in terms of our understanding of the technology and our adoption of the technology. And when I wrote that note to myself, I was like, wait a second.

[00:42:34] Didn't I write an article about this? And so that led me back to March 2nd, 2023. So this is three and a half years ago now. I published the Law of Uneven AI distribution. So this was two weeks. Before GPT-4 came out four. Yes. So this is right after chat. GPT was launched right before GPT-4, [00:43:00] and so I'm just gonna read the beginning of this because it's basically like copy, paste and insert super intelligence and like agents.

[00:43:08] So the law, that's how I started it. The value you gain from AI and how quickly and consistently that value is realized is directly proportional to your understanding of. Access to an acceptance of the technology. So then I continued. I said, I've been thinking a lot lately about the AI adoption curve, both in our personal and professional lives, and who stands to benefit most from rapid advancements in the technology.

[00:43:33] Where I've landed is that the impact and benefits of AI will be unevenly distributed to individuals, companies, and industries. In some cases, it will be your own choice. And in others it will be by institutional design. For example, financial services companies blocking employees ChatGPT access, which is still happening three years later.

[00:43:52] Educational systems at all grade levels struggling to adapt to AI capabilities and curriculum still happening three years later. [00:44:00] your own unwillingness to install super interesting AI apps because you don't know or trust the companies and how they use your data. Still happening now with agents like turning 'em on or not turning on.

[00:44:11] I said, this uneven distribution will create dramatic differences in people's experiences with and perceptions about ai. It will profoundly impact how much you reap the benefits of AI in your professional and personal lives. How much value your company extracts from the technology. And the trajectory of your AI journey.

[00:44:30] and then I said factors that influence distribution of benefits, they're the three primary factors, understanding, access and acceptance. So understanding is like just situational awareness. Like what is the technology, what is it capable of? Deep comprehension of that. And then your competency of using it.

[00:44:47] Those are still a hundred percent true. Like we see that all the time. Access to it is like, are you allowed to use it within your organization? Are you, are there gatekeepers keeping it from you? Are the labs themselves keeping you from [00:45:00] accessing the most recent technology, the most advanced models?

[00:45:03] And then the acceptance of is like, what do you have to give up to get it? Agents has become a huge part of this. It's like, hey, if you wanna get the full benefit of agents and really see what's possible, you gotta let 'em be proactive. You gotta give 'em access to your life and then they, all of a sudden you see it.

[00:45:16] And so that's what I took away from this Mike, is like. We're just increasingly going to be living in this world where the models are getting way smarter, their capabilities are expanding exponentially. There's going to be this huge impact on people and companies, but not everybody's gonna experience it in the same way.

[00:45:33] And some people are gonna be oblivious to what's going on. And that, that to me is like what we see. And so now it's just like, it's like whatever we were thinking in 2023 is just accelerating. 'cause now you have the, you know, on the cusp of super intelligence, you have agents that are reliable and autonomous in certain settings and it, and then you still have companies that are sitting around trying to figure out how to use AI assistance.

[00:45:55] Like, it's just a weird dynamic right now where like. The, [00:46:00] the people understand it are moving forward. Everybody else is kind of sitting back like, having no clue this is happening.

[00:46:05] Mike Kaput: Yeah. It's such a weird time when you kind of see both sides of that, you know, I'll, I'll confess, I sometimes find it hard to give developments like AI 2027, their due weight, because like, I personally, I think understand better than many people what this could mean for the near future trajectory of AI and business.

[00:46:27] Or at least have suspicions about where we're headed. But I also then, like anyone else, have priorities this week, getting the tools we already have to work better for us. And that tension is hard to navigate.

[00:46:39] Paul Roetzer: Well, and you know, this kind of goes to the whole AI Academy Day on Thursday. Yeah. I like, this is my struggle.

[00:46:46] Like we have, you know, if you think about it. Especially within our listener audience for the podcast. You, most people in this audience are, are probably in like that one percenter, like, you know, top 5% for sure. In terms of their [00:47:00] comprehension, their probably access to like their understanding, their access to it, their willingness to like give up some, you know, some privacy to like get the benefits of the tools.

[00:47:09] They're racing their companies forward, like they're doing all the things and they're roughly situationally aware to like where this is all going. Yeah. And then you have the rest of the corporate world and the SMB world who maybe they have licenses to the tech and maybe they're using it to like, as a thought partner, maybe it's like doing some emails and some meeting summaries and like.

[00:47:29] They're just using it like any other tech and like how do you provide an education platform that serves both those audiences? Yeah. Like this is what I've, I've thought deeply about this for a long time, but more so in the last, like six months. Yeah. and so this is where I'm really thinking, and I think we have some solutions for it, but it's an interesting environment right now with how fast the tech is moving.

[00:47:53] Mike Kaput: It really is. All right, before we move into rapid fire, this episode is also brought to you by AI [00:48:00] Academy by SmarterX. So you hear us talk about AI Academy all the time, but there has never been a better time to become a part of it because we are having kind of a special offer throughout November. that kind of was kicked off, celebrating our 250th podcast episode.

[00:48:16] So the offer is if you join AI Academy by November 30th, 2026, you can get $250 off a new AI Academy individual license if you use the Code Pod two 50 at checkout. This is for individuals, so that's the full. Academy experience, courses, certifications, live sessions, everything we're building that you'll hear about, especially on Thursday, to help your, you and your team get ahead with ai and that's 25% off your first year, 250 bucks off if you're a business leader charged with driving AI literacy and transformation for your team or organization.

[00:48:54] We also have an offer for you, which is if you buy 25 or more licenses by November [00:49:00] 30th, Paul will do a private one hour virtual session with your team. If you buy 250 or more licenses, Paul will come to you for a US only private in-person session in the first half of 2027. So there has never been a better time.

[00:49:18] If you are considering joining AI Academy or getting your organization started with a business team account, this is the time. Head over to Academy dot SmarterX dot ai for an individual. Use the code POD two 50 at checkout to get that discount. If you are. Talk in team licenses. Reach out to our team to talk through options for your organization.

[00:49:40] This offer runs through November 30th

[00:49:43] Paul Roetzer: and again, join AI Academy Day. Like if you're in, if you're not sure, yes, it's the right fit for you, come and check it out on AI Academy Day and then you can make make a call from there. But yeah, I, like I said, there's some pretty cool things we're working on. I think it's gonna be a really good time for people to jump in.

[00:49:58] Fable 5.2, Mythos 5.2, Gemini 4 Carbon, and Gemini 4 Hydrogen

[00:49:58] Mike Kaput: All right, so let's dive into [00:50:00] some rapid fire. So first up, this past week we got a ton of new models, Anthropic release, Claude Fable 5.2, and, in, or announced at least, mythos 5.2 while Google expanded access to Gemini four carbon and introduced Gemini four hydrogen. So starting with Fable 5.2, Anthropic positions this as above Opus.

[00:50:20] 5.5 for ambitious coding, research and engineering projects. This update improves reasoning, visual understanding, and the ability to keep working through complicated tasks. In one experiment, Anthropic reported Fable coordinated more than a thousand agents over eight days to build a prototype operating system that boots on a laptop.

[00:50:41] Anthropic also used it to design circuit boards, enclosures, and software for three electronic devices that humans then manufactured and assembled. Mythos 5.2 uses the same underlying model as Fable without certain cybersecurity and biology safeguards. Direct model access to Mythos [00:51:00] is offered only through the verification program that Anthropic has for approved security and life sciences work.

[00:51:06] That includes tasks that Fable would block. So this model is not available to everyone. Anthropics says Mythos is better at finding, reproducing, and fixing software vulnerabilities in biology. It developed a method for modeling protein. Shells used to deliver gene therapies accurately modeling 28 of 34 structures in the company's test.

[00:51:26] And both models generate output up to 30% faster than their 5.1 predecessors according to Anthropic. So if you're a. General Claude user Fable is what you're gonna be using or checking out. It's available on paid Claude plans. Google on the other hand, is expanding Gemini 4 Carbon to general release for users.

[00:51:47] We talked on previous episode, they had introduced it. It was a limited release at the time. This is designed for difficult work that takes many steps such as ma, major software changes, analyzing dense documents with charts and [00:52:00] engineering drawings, and producing spreadsheets, presentations, and reports.

[00:52:04] Carbon is now available through Google's API. It's rolling out to Google AI Ultra subscribers. It's in AI mode in search at the moment, and they're moving it into general release as we speak. Last but not least, Google also introduced Gemini four hydrogen. This is adapted from carbon. It is for faster everyday coding, research and writing in Google's tests at the highest high reasoning setting.

[00:52:29] It completed tasks two and a half times faster than Gemini 3.8 Flash. It can make targeted code changes, run tests, and combine info from documents, spreadsheets, audio, and video. Hydrogen also powers a search upgrade that breaks complex questions into simultaneous searches in Google search and returns more visual comparisons and relevant links.

[00:52:51] So Paul, tons of new models that would've been a loan worthy of a whole month of updates. I feel like in the past. this is the [00:53:00] new rate of acceleration. I will just say I. Feel like, I confess, I have to go back to science class to remember how the hierarchy of Google's models works. I feel like an idiot looking at this, honestly.

[00:53:11] It's just,

[00:53:11] Paul Roetzer: I'm literally, as you're talking, I'm in Gemini Enterprise asking this question 'cause I, oh my God. Yeah. I also feel like an idiot. So if I go into Gemini right now, and I, so again, I'm in enterprise, so you might, if you're in a regular Gemini license, you might have all these models. So when I click the, which model to use in the little dropdown in the chat, it pops up auto, where would they choose the best?

[00:53:35] I then have four hydrogen, four carbon 4 Argon, 3.8 flash. 3.7. I have all the flashes and I still actually have 3.1 pro, which is in preview mode, which is weird. And then 2.5 pro. So I said, what is the theory behind Google? Hydrogen carbon argon? How are we supposed to know which is stronger? it says, Google's new naming scheme, for the models, represents a [00:54:00] shift in how they name and categorize AI capabilities, moving from labels like flash and pro to elements on the periodic table.

[00:54:05] the theory behind is how to determine which is stronger, is detail. Below it then says hydrogen represents the lightest element. Argonne positioned as the middleweight noble gas and carbon positioned as the foundational organic element. I was like, okay, whatever. And then I like scroll down. So then I said, is the average person supposed to know by elements which is better?

[00:54:25] And it says no. The average person is absolutely not expected to look at these chemical elements and instantly know which is better. Historically, tech companies used like clear base. By switching to this, they've swapped intuitive grading. For a metaphor that requires a chemistry refresher to decode, that's usually a bad sign.

[00:54:42] Here's the cheat sheet though. The atomic weight cheat sheet. How the metaphor works. Oh my God, oh my God. To make sense of names, the community has adopted the atomic weight rule. Heavier elements on the periodic table represent larger, more computationally heavy and more quote, intelligent models. So hydrogen is atomic [00:55:00] number one, and I'm not, I have not verified this.

[00:55:02] I don't have the periodic table up while I'm looking at this. So like, check yourself. so hydrogen, the atomic number one, the absolute lightest element is the lightest model. Pure speed, low cost, fast execution. Carbon heavier than hydrogen is the flagship, highly structured organic element, representing the deepest reasoning and heaviest intelligence.

[00:55:20] And then argon physically heavier than carbon. But in Google's lineup, it sits as the noble gas middle. Oh my God. So what if you're confused, like welcome to the club. This is what happens when probably the engineers like sat around and were like, dude, you know, it'd be sweet. We did at the P table. Right.

[00:55:40] And I remember back Mike, like you and I spent some time at this bar at MIT, we used to hang out at HubSpot. There's this bar at MIT where the menu is, the periodic table. I dunno if you remember that bar. And so it's like, this seems genius to like really smart people probably. But to the average user, you're like, what the hell am I [00:56:00] supposed to do with this?

[00:56:01] And I mean, I literally will have to probably have a cheat sheet like reminding me which one is like. The right model to use for like what I'm trying to do

[00:56:07] Mike Kaput: and yeah, I give them credit because we have talked about like, okay, get more interesting names. Yeah. So it's like, yeah, it's cool. But like if, even if you go into Gemini right now, it's like, and you look at the, not Gemini enterprise, but just Gemini, the web app and you look at the models, it's like, okay, 3.5 Flash-Lite.

[00:56:26] It says it has like helpful fold labels, quick answers. Okay. Four hydrogen everyday help, 3.1 pro detailed research. I don't know why four hydrogen wouldn't be better for re research given the gap in models. Four argon code and analysis. Okay, but isn't. I don't know even what code fine analysis, isn't that also included in research or every day?

[00:56:50] I don't know. For carbon, which is locked to me, is, I don't have the ultra account at the moment, most complex work. But like, okay, like what does that mean? What does that mean? And none [00:57:00] of these labels are that helpful. And maybe it's an impossible task. I don't wanna bag on Google, correct. All the labs have this problem, but at least like OpenAI's hierarchy.

[00:57:09] I like, I'm like, oh, okay. The moon is smaller than Earth. So like, Luna is smaller than Tara. And even that's like, yeah, you gotta think. Yeah, yeah, yeah.

[00:57:19] Paul Roetzer: And Opus and yeah, haiku and like, oh

[00:57:21] Mike Kaput: my God, those are, those are hard too. It's just, and it's like kind of interesting cool names. So I'm like, my God, this is an order of magnitude too complicated.

[00:57:29] Paul Roetzer: and my theory is, and I don't, I have no idea if this is right, is. The marketing teams probably play almost no role in any of this. And so you have, and they probably get these models dropped on them like three days before the announcement is being made. Like, Hey, we built these amazing new models.

[00:57:48] Here's the names for them, here's what they do. And the marketing team's like, what? And then they go into their own AI model and say, help us understand what the, you know, the research team is saying, like, this makes no sense. and then they slap on these [00:58:00] like simple things. yeah. It's,

[00:58:02] Mike Kaput: yeah.

[00:58:02] Paul Roetzer: I don't, I’m shocked actually that four years later we are still in this state where they, they just can't get out of their own way.

[00:58:09] Mike Kaput: Yeah.

[00:58:10] Paul Roetzer: And I do think the answer's eventually, like, it, it, for 99% of users, there is no model choice. Is, is like the likely outcome of all of this. Like, you just go in and just talk to it and it, but I thought that three years ago, and here we are.

[00:58:26] Mike Kaput: Yeah. Yeah. All right.

[00:58:28] Brian Halligan Returns to “Refound” HubSpot for the AI Era

[00:58:28] Mike Kaput: Some other big news happening this week.

[00:58:30] So this past week, HubSpot, co-founder Brian Halligan, returned as CEO of the company saying he and co-founder Dharmesh Shah are refounding the marketing, sales, and customer service software company for the AI era specifically. So he is replacing Yamini Rangan, who had led HubSpot since 2021. As CEO Halligan says, the overhaul will reach every aspect of the company's product, sales approach and culture.

[00:58:57] Their new direction from what they've been [00:59:00] posting online and in statements are what the founders call a quote customer relationship harness. So Shaw says HubSpot is shifting significant resources towards making more of its capabilities accessible to AI assistance like Claude and ChatGPT. That seems to imply you may soon be able to do more of your HubSpot work just by asking AI assistance to go do it for you, the ones you already use, not the ones within HubSpot.

[00:59:23] And then HubSpot supplies customer information and perhaps tools and capabilities behind those actions. So Halligan return comes as the business faces some serious pressure. Bloomberg reported HubSpot stock has fallen about 50% during 2026 so far. Amid slowing growth and concerns ai, native comp competitors could disrupt its business.

[00:59:45] Halligan told Bloomberg that HubSpot may have focused too much on short-term financial targets and too little on building great products. He said he wanted to keep HubSpot from shrinking or ending up owned by private equity and instead rebuild and reimagine the company. So, [01:00:00] man, Paul, I mean, very few people on the planet have a better, are better qualified to talk about this than you given your.

[01:00:06] Background with HubSpot. What was your reaction to this? Surprise?

[01:00:10] Paul Roetzer: yeah, I mean, definitely surprised. So for context, we were HubSpot's first partner. So my agency was their first partner back in 2007, I think we started working with them. so we came on as a customer in oh seven and started reselling HubSpot software in oh eight, like two or three years before they actually created a formal partner program that Halligan sort of green-lighted with our friend Pete Caputa, who we talked about recently as a data box, CEO.

[01:00:35] Now, yeah, a little context. I guess like, given my background with them, have a, a few insights. so Halligan. Post on X was, I'm back as CEO, HubSpot. Today, Dharmesh and I are completely refounding the company. It's day one, we're now the customer harness company. I'll come back to that in a minute. In the next few days, I'll post about why I am coming back and what my plans are.

[01:00:58] Let's go. The [01:01:00] refounding HubSpot is going to take is where the puck is going. That was always one of Halligan's, Halligan's favorite sayings in the early days of HubSpot.The new tagline, the customer relationship harness, HubSpot's harness is on fire. HubSpot inside Claude is on fire, pouring gas on the fires.

[01:01:18] And this almost feels like Mike, this feels like the naming of AI models. To me, the customer relationship harness means literally nothing to anyone outside of the technical world. Like,

[01:01:32] Mike Kaput: no,

[01:01:32] Paul Roetzer: I can't imagine. That anyone in the HubSpot ecosystem other than technical partners know what a harness is.

[01:01:40] Right? You and I talk about harnesses. Yeah. And I sometimes still struggle to actually explain what is meant by it. 'cause I think people mean different things. And then the other thing they're leaning heavy into is headless. Now we talk about headless on this. Again, I don't think the current HubSpot customer base understands what headless is, [01:02:00] nor will care what headless is.

[01:02:02] And I actually saw Dharmesh comment on somebody's LinkedIn post saying he fought against the use of headless in the messaging, but kind of lost because the market was already there. And it's like, I don't, I don't agree with that. Like I think the technical market is there. I don't think the HubSpot customer base, which is largely SMBs.

[01:02:22] Is there for either of these things. Harness or headless now. Yeah. Doesn't mean it couldn't work, but yeah. So yeah, my history with Brian Dharmesh, they're, they're friends of mine, we go back to 2007, spent a lot of time with them in the early days of HubSpot. Had the privilege of sitting in a lot of like really amazing meetings in the early days of HubSpot.

[01:02:39] I thought it was interesting that they called it in the press release, they said HubSpot is the the agentic customer platform for scaling companies. That's how they defined it. So that was just an interesting note. And then Dharmesh posted as part of the HubSpot Refounding. Halligan and I are working to get back to these core things.

[01:02:59] Easy to buy, [01:03:00] easy to use, easy to love. Those were three things. And then that led me back, this is like walk down memory lane day for me, going back to all these posts I wrote. So back in July. 2013. So this was a year before the HubSpot IPO, they were, they were obviously heading towards the IPO. And so I wrote this article called Why We Bet On HubSpot six years ago, and Why You Should Now.

[01:03:24] So again, this is me owning my agency, being a top agency partner of HubSpots at that time. And I think it's relevant. So I'll just kind of read this and then we'll, we'll move on with the day. So I said back before the a hundred plus million in venture capital Cash, the inevitable IPO, the European expansion, the 9,000 plus customers and the 900 plus certified agency partners.

[01:03:44] The rise of the inbound marketing agency, which is the term I coined back in like 2007, the Birth of the Sales Lion, which is Marcus Sheridan, who had become like a big partner at that time. And the introduction to some of this tech. there was a small team of marketers with a vision to transform an industry.[01:04:00]

[01:04:00] Like any high growth company, there have been bumps in the road. As HubSpot's first agency partner, we had a front row seat through the years. There have been times when I questioned product decisions, customer communication strategies and agency partner program moves, but that's never swayed my belief in the people and the vision.

[01:04:16] HubSpot's transparent almost to a fault with their employees and customers, which they're doing now. Like they are live tweeting the refounding of this company. Yeah. when they fail, they own up to it and they do everything in their power to make things right. Why? Because they actually care. They believe in their cause, and they know trust is essential for customer happiness and retention, which are critical for their success and perpetuation.

[01:04:39] As Brian has said, on countless occasions, they're in it to build a hundred year company. We partnered with HubSpot in 2007 because I believed in the people and the vision. Starting at the top with Brian and Dharmesh. It's a company filled with innovative minds and intrinsically motivated professionals.

[01:04:55] They believe in their cause and their customers. And for nearly six years at that time, I'd witnessed [01:05:00] firsthand the impact of HubSpot. I will say I feel like they drifted from that for a while. You know, we're just a HubSpot customer now. I sold my agency five years ago, like I have no stake in the game.

[01:05:10] but as a customer even, you just feel a lot of friction in the community. But I think it's worth sharing how I felt about Brian and Dharmesh back then, and I, that hasn't changed. Like I still believe deeply in Brian and Dharmesh, and I'm, I'm rooting for them. I hope they figured that out. They figured out the future.

[01:05:28] Like it's a hard problem to solve what they're doing many SaaS companies face right now. And I admire that they're willing to take a swing at this and dive back in. They have no financial reason to do this. Like they, they both could be retired and just spend their time investing and not be in the game.

[01:05:48] And so, you know, I think it's admirable that they're willing to take a high profile swing and it might not work, but I will be rooting for them. They're, they're good people. and I believe that their heart is [01:06:00] in the right place with what they're trying to do.

[01:06:02] Mike Kaput: I'll be following the live tweeting of the refounding with great, great interest.

[01:06:06] I mean, regardless of how it goes, it feels like they're tearing the bandaid off that these SaaS companies basically need to tear off, right? Because you have to just go all in on the fact that in some way, you have to go all in on the fact that you're not making AI as good as the frontier models.

[01:06:21] Paul Roetzer: Yeah.

[01:06:21] So, and it's gonna be messy because like right now it's, I don't wanna say easy, but like the branding, the taglines, the, you know, culture code, that's the nice stuff to tweet. Your current team isn't the team that's gonna do this. Like there's, there's gonna have to be massive turnover. There's gonna be massive change in org charts and structure and

[01:06:44] It's gonna be messy and that, that's not the fun part to do. But they obviously are, aren't gonna hold back. They're gonna do what they think has to be done. And with that is gonna come some pretty significant change in the company, the product, the people, all of it's gotta [01:07:00] hit the reset button if you're gonna do it.

[01:07:03] SaaStr Sets New Revenue-per-Employee Baseline

[01:07:03] Mike Kaput: Next up, Saastr founder, Jason Lemkin, we've talked about a bunch, has published a new analysis arguing that companies need to rethink how much revenue per employee is now possible thanks to ai. So revenue per employee company revenue divide by headcount. Lemkin argument is basically the AI lets a business support more customers and more revenue without expanding its sales, marketing, and back office teams at the same rate as before, he cites this benchmark from 2025 by a company called iconiq, where they said, look, it's, there's an average of $237,000 in annual recurring revenue per full-time employee for a typical B2B company.

[01:07:41] In its newer group of fast growing AI focused companies that they studied with at least a hundred million dollars in a RR or revenue, the median is now $655,000. Of revenue per employee. Anthropic, for instance, illustrates how revenue can outpace hiring. They report $6.6 [01:08:00] million per employee for the year through September.

[01:08:02] That's up from $1.4 million a year earlier. So Lemkin basically describes this is happening. This is how it works inside SaaStr. He says their business is on track for $20 million in revenue this year with three people that I think that's Saastr specifically and roughly 20 AI agents. So they have gone all in on agents.

[01:08:23] We've talked about that quite a bit. agents handle prospect outreach, follow-up renewals, invoice invoicing, collections. People do customer replies, unusual pricing, major relationships. They check the agent's work. Lemkin even describes these recently spent a 13 hour day just reviewing and fixing what agents had done.

[01:08:41] So basically he's saying, look, the game has changed and the companies generating less than $300,000 per employee and growing below 30% should expect questions from their board. So Paul, I'm curious from your perspective, obviously anthropics business aside, that's a pretty unique and incredible business, [01:09:00] but it sounds like expectations around revenue per employee are rising.

[01:09:03] Do you agree with that?

[01:09:05] Paul Roetzer: They should be for sure. Yeah, I mean it's definitely one of the factors, like when we build our AI value calculator, which, we'll, we don't really talk about much, but it is on the site. It's a free tool. We can throw it in there. It's one of the things that was core to the premise of like, you should see this increase in revenue per employee.

[01:09:20] Saastr is an extreme example. I've said on the podcast before, they are doing things that I would never do. Yeah. Like, I would never be willing to take the risk on that they're taking on. but it's certainly fascinating to watch from the sidelines. Interesting enough, I'll drop a link in the show notes.

[01:09:39] Jason had thoughts on the refounding of HubSpot, so just to tie the two together. He published a, a post, this was I think on Sunday. He said HubSpot is refounding as the customer harness company. Here's what that means for us right now. In November, 2026, he said HubSpot just named, what we've been doing at Saastr for the last year.

[01:09:58] we are [01:10:00] 100% headless on our CRM. We left traditional marketing automation entirely in the last 60 days. Reading Halligan's announcement. Felt like reading our own roadmap, published by one of the biggest companies in B2B. what does customer harness mean in practice Right now? We all have different takes.

[01:10:15] we're at different stages, but then he goes through, we are 100% headless in our CRM and never going back. We moved off traditional marketing automation. Claude ChatGPT and Slack are great, but our main surface is 10 K, which is like their agent that they built. We use the best agent for the job, whether or not it comes from our CRM vendor, there is no agent loyalty.

[01:10:33] But it's hard to get us to look at other agents once we're happy and and then it just kind of breaks down the kind of where they're going that's all tied to their need for so few people in the company.

[01:10:43] Mike Kaput: Hmm, interesting.

[01:10:44] Paul Roetzer: Yeah.

[01:10:46] United Airlines Moves to Block AI Agents

[01:10:46] Mike Kaput: All right, so another kind of evolving story. This past week United Airlines announced they are imposing restrictions on third party AI assistance and agents that book flights or manage travel for customers without the airline's permission.

[01:10:59] So these [01:11:00] restrictions cover United's website and app plus access through travel agencies and booking aggregators. They're updating their book booking, ticketing and distribution policies. They basically say sharing account credentials with a personal AI agent or assistant can expose sensitive information or lead to unintended purchases, account changes or the use of MileagePlus miles.

[01:11:21] They warn that automated actions could violate FAA or transportation department rules. and they also say that approving an agent's action yourself doesn't get around their restriction. Their website terms require the airline's own written permission, including for agents that can search that search its site.

[01:11:40] Travelers can still use AI for general trip research, provide the tool doesn't access United's website or app. I don't know how you'd stop it from doing that, but, United also says agent made bookings from before November 11th won't be canceled. They say they want standardized secure integrations with privacy authorization and accountability [01:12:00] protections in place.

[01:12:01] They currently have no partnerships with third party AI agents. So Paul, like, is it just me or does it seems insane to me that they think they can restrict and regulate this. Like I understand why they want to, like, there's no question they bring up like some interesting points, but like, given how fast this moves and how much people are trying to use agents for things, this seems wild to try to figure out.

[01:12:23] Paul Roetzer: So first I wanna flip this and say, as a user of agents who I'm sure would be sending your agent to United, how does this make you feel?

[01:12:32] Mike Kaput: Yeah, I am upset, I would say, or I'm actually just more curious than upset. Okay. But I'm just, and actually I'll mention that United, it's funny. United is the example yours that comes into the use case segment.

[01:12:46] Okay. All right.

[01:12:47] Paul Roetzer: Good.

[01:12:47] Mike Kaput: But yeah, it's, it's just weird. I'm just like, okay, like a, how do you enforce this? I don't even know. I think you can kind of do it, but also B, it's like, why, like this is the future. Yeah. I'm gonna be asking Dot or whoever [01:13:00] to do stuff for me.

[01:13:01] Paul Roetzer: So I actually had a thought while you were reading this.

[01:13:03] I had, 'cause I haven't honestly spent too many brain cycles actually, like contemplating this whole, like, shutting down agent access to things. But as you were saying it, I, the thing that immediately came to mind is like, oh, this is just leverage to, to charge them. Yeah. So you, you're basically a gatekeeper.

[01:13:18] So if an agent can identify itself as coming from OpenAI and OpenAI is paying United a billion dollars a year or whatever to like do it, or they're getting a percentage of the booking, I. That's like, again, I may be completely wrong on this.

[01:13:31] Mike Kaput: Yeah.

[01:13:32] Paul Roetzer: But it seems like we're just in a leverage game now where the gatekeepers are trying to shut it down to like.

[01:13:39] Ensure the security slash make some money off of it. I don't, I don't know. I don't think two, three years out it makes sense that people are like blocking agents. Yeah. But maybe there's like verified agents are allowed through in some way.

[01:13:53] Mike Kaput: I'll be honest. I hope that's the logic, because if their logic is simply we're gonna build our own AI and it's [01:14:00] United's crappy.

[01:14:01] Yeah, exactly. Like I'm gonna be furious about that. Yeah.

[01:14:04] Paul Roetzer: I That might be the play too. Maybe they do just think that they can just wall,

[01:14:07] Mike Kaput: that's wall

[01:14:08] Paul Roetzer: garden

[01:14:08] Mike Kaput: approach. That's crazy. That's not gonna work.

[01:14:10] Paul Roetzer: Yeah. That would not be headless. That would not be, that

[01:14:13] Mike Kaput: would, right. This is they can't

[01:14:14] Paul Roetzer: build their own harness.

[01:14:16] Mike Kaput: Right? Right. Exactly. Yeah. Alright, moving on.

[01:14:19] GPTs > Plugins Update

[01:14:19] Mike Kaput: So next step, we've covered OpenAI's decision to retire custom gpt. So those customized versions of ChatGPT built with instructions and reference files for specific use cases. The standard retirement date is December 11th. OpenAI is recommending people move those workflows into plugins, which combine reusable instructions in the form of skills, reference material, and connections to other apps.

[01:14:42] So we wanted to revisit this because we're actually working through this transition with our own public facing GPTs at SmarterX. So Paul, I just wanted to kind of open it up and just kind of get the latest on what we've been thinking. I know we've both been involved in this to some degree.

[01:14:56] Paul Roetzer: Yeah, I think just like a heads up for everybody.

[01:14:58] So the way we are doing it is we we're [01:15:00] gonna move 'em to plugins. It only solves for people in the, in the ChatGPT ecosystem, you know, you don't be be able to use that. so we're simultaneously still looking at ways to, support people like for jobs. GPT in particular is the one that's like extremely popular, but people also use our innovations, GPT and Problems.

[01:15:16] GPT. Campaigns, GPT, we, we incorporate those into our courses and stuff. So we're gonna rebuild those as plugins so people will be able to go to the same URL on our website and get access to them as plugins. And then we, we will likely include some version of the prompt where it's like, hey, if you want to just go do this yourself, like here's, you know, you could use this prompt and get roughly the same output from your platform of choice, which is what we do in workshops when we're running it.

[01:15:41] Yeah. It's like, hey, if you're not on, if you're on copilot and you still want to use this function, here's a prompt you can use. So, yeah, just kind of a heads up, we plugins are seem like the best option. It's where they're forcing you to sort of go does, it's not a one-to-one comparison, like apples to apples, but it seems like it's gonna function as needed.

[01:15:59] Mike Kaput: [01:16:00] Yeah.

[01:16:01] AI Use Case Spotlight

[01:16:01] Mike Kaput: Alright, so onto our AI use case spotlight, where every week we give you a look under the hood at some real AI use cases we're exploring. So I'll be super quick Paul, and I want to hear your experience. So I have been experimenting very, very, in a novice fashion with Dots from OpenAI. So. I actually, for me, the big unlock here was I had the ability to connect my text messages to so I can actually text dot.

[01:16:26] It's a contact in my phone now. That was super helpful. I started using it, trying it out for a bunch of different random personal tasks, one of which was trying to get a refund from United for something. So I texted DOT and said, Hey, go look at my, like notes on this. I'm basically have this travel credit I need to use by a certain date.

[01:16:47] I'm not gonna be able to use it by that date. Can we go try for a refund and dot figured out how to go chat with United Support, get a few things resolved. I'm still waiting on a couple things, but the ball was moved forward by [01:17:00] dot, which I guess United did not have that locked down in whatever way Dot decided to do that just yet.

[01:17:07] I did not use it to book a flight, like, don't come after me, you know, FAA or United, or whatever. True. I used it to chat with support, but that is interesting to just even consider that use case and like, oh, it's like going off and doing all this stuff and reporting back while having a reasonable, useful conversation with another human or bot.

[01:17:26] I'm not sure. And then just bringing the result back to me was extremely valuable for me as something I had put off doing because I didn't wanna do this myself.

[01:17:38] Paul Roetzer: Yeah, mine was related to dot and actually it was a weird one. I dunno if you saw this, Mike. I posted this in the internal channel yesterday, but, so I'm sitting at the, I was at the, actually the Browns game on Sunday and I get an alert from my, my personal dot and it said, quick heads up before Thursday's, AI Academy Day.

[01:17:56] And it, and then it made some suggestions to our. [01:18:00] Business account page. and then it like actually suggested some fixes to the academy page. It's like, here's some changes you might wanna make to the page. And I was like, what the hell is this? Like, so again, proactive agent stuff, but this is my personal agent.

[01:18:15] And so I replied to it. I said, what in your memory base? Based on our chats triggered that suggestion. Was there a specific chat where I talked about that? Like AI Academy Day? How did you know it was happening? Because I don't have conversations in my personal ChatGPT account about this stuff, and it said no specific chat triggered it from our past chats.

[01:18:37] I knew you lead SmarterX and that AI Academy is a priority for you. That's why finding the findings seem worth flagging, but it's not how I knew the event was happening. A background research check surfaced Academy Day. On the Public Academy homepage, which lists November 19th. You also mentioned it on episode two 50.

[01:18:58] So my dot apparently [01:19:00] is like monitoring the transcripts to our podcast. I don't know. And I said, I checked those public pages myself and noticed the conflicting copy. I didn't use your calendar or your email. Wow. I was like, what the hell? So I still don't know if that's actually true, but if my.is proactively running searches on me and like my public calendar, like.

[01:19:25] I don't know. I haven't really processed this yet. and then I also used, QuickBooks. I love the plugin for QuickBooks and ChatGPT on my business account. So I have conversations all the time with like financial related stuff. And then I'm working on a, a big project that had a bunch of like, legal documents and everything, and I was like, everything's just going in there.

[01:19:42] What do I need to know about this? So I need to say to the attorneys. so yeah, I just, I had a bunch of like, in little use cases and then some proactive stuff with my Dots, apparently. Oh, and Dots are now in your business account. That came out on Sunday. We now have Dots in our, and then HubSpot Home appeared in my ChatGPT business account.

[01:19:59] I don't know what that [01:20:00] is yet, but

[01:20:00] Mike Kaput: yeah,

[01:20:01] Paul Roetzer: it looks llike a formal integration with HubSpot. I guess, I don't know.

[01:20:05] Mike Kaput: Interesting.

[01:20:06] Paul Roetzer: Says it's in beta. I haven't tried it yet.

[01:20:08] AI Product and Funding Updates

[01:20:08] Mike Kaput: Okay. Alright, we're gonna wrap up here with some quick AI product and funding updates. I've got just a laundry list of these. I wanna get through Paul to kind of close out the episode.

[01:20:19] So first up, OpenAI began rolling out something called ChatGPT, wallet to us ChatGPT Pro Subscribers through its Dots agents allowing agents to complete purchases with user approval and security checks for each transaction. OpenAI also launched something called Businesses in ChatGPT. This is in beta.

[01:20:38] This lets companies claim and manage profiles and product feeds and see engagement data in ChatGPT. So basically you can claim your business and start seeing how people are engaging with it. eligible US e-commerce and online service businesses can also launch agents that answer customer questions.

[01:20:57] OpenAI's third quarter revenue rose about [01:21:00] 50% from the previous quarter to $10 billion. Operating losses widened to $10.9 billion according to the Wall Street Journal. OpenAI and Upwork together did some work where they found that activities common in chat. GPT tended to lose share of Upwork postings.

[01:21:15] It's not super surprising. While AI related projects carried larger budgets and involved more hours of work, Upwork being a freelancing platform, so interesting data on where that market is. Anthropic introduced Claude meetings and beta for team and enterprise customers that lets Claude join Google Meet calls to take notes, answer questions, and create documents.

[01:21:35] During the meeting, Anthropic also launched collaborative spreadsheets and whiteboards in beta for paid Claude users and made its dashboard tool generally available as something called Claude Data. Anthropic expanded interactive connectors so users can open supported apps inside Claude and have the assistant make changes directly in those tools.

[01:21:54] They also made its unified Claude experience generally available to team [01:22:00] customers. This combines chat and Cowork will be talking about this, I'm sure, moving forward so users can ask questions or delegate longer tasks from the same conversation. Instead of two different areas, Google has opened pre-orders for Gemini powered glasses from Warby Parker, starting at 395 bucks, arriving December 10th.

[01:22:21] This has hands free assistance based on what the wearer sees. Google's Gemini Notebook can now answer questions and generate content from users' own notes alongside imported sources granola, the of AI voice tool launched agent briefs for business and enterprise users that turns meeting history into nightly summaries of the work and priorities that connected assistance such as Claude and Chad GPT can use.

[01:22:48] Apple is reportedly exploring a more autonomous Siri that could track user goals and pursue tasks without step-by-step prompting. According to Bloomberg and the Center [01:23:00] for AI safety reports that Fable 5.2 completes 41.3% of remote labor index freelance projects, which is a benchmark used to see how good AI is at that type of work at a professional quality versus 2.5% for the best system when the benchmark launched a year ago.

[01:23:21] So sounds like we've made some significant progress and just as one final reminder for folks go to SmarterX dot ai slash pulse this week to take this week's pulse survey, we are asking about how you would describe the current pace of AI improvement in your own work. Super excited to see the results of that Paul.

[01:23:41] Thanks for breaking down a wild week. We'll be watching Jensen Anthropic and the rest pretty closely this whole week.

[01:23:47] Paul Roetzer: Absolutely. And do we have an episode next week, Mike? I can't remember, like we've got next week, the holidays.

[01:23:51] Mike Kaput: We, yeah, we're gonna still release one. Normally you and I are recording a little early on that one.

[01:23:56] Okay. But yes.

[01:23:57] Paul Roetzer: Okay. So we will have one more episode, before the [01:24:00] Thanksgiving holiday. That's, it's next week. Thanksgiving.

[01:24:02] Mike Kaput: Yeah. Next week. Yes. It's,

[01:24:05] Paul Roetzer: oh my gosh. How is that possible?

[01:24:08] Mike Kaput: That's a great question. We'll also have an AI transformations episode this week as well. Okay, cool. Hopefully, fingers crossed.

[01:24:15] As long as the recording, stays for tomorrow, you know, knock on wood. If not, it'll be after Thanksgiving, but we're moving forward with that series as well. So if you haven't seen an episode for that in a while, don't worry. Coming. Those are great.

[01:24:27] Paul Roetzer: I hear from people all the time that they love that series.

[01:24:30] Mike Kaput: So that's, yeah, it's cool to do.

[01:24:31] Paul Roetzer: Alright, well thank you everyone. Have a great week and we will talk to you again next week. Thanks for listening to the Artificial Intelligence show. 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 earn professional certificates from our AI Academy and engaged in the SmarterX [01:25:00] slack community.

[01:25:01] Until next time, stay curious and explore ai.

Recent Posts

[The AI Show Episode 252]: Jensen Huang Resigns from Nvidia, Anthropic’s Mega IPO, AI 2027’s New Warning & HubSpot’s AI Reboot

Claire Prudhomme | November 17, 2026

Jensen Huang resigns from Nvidia, Anthropic launches its IPO roadshow, and AI 2027 authors accelerate their forecast, Ep.252 of The Artificial Intelligence Show.

[The AI Show Episode 251]: Forward-Deployed Engineers, Shadow AI, the Owner Bottleneck, AI on Your Resume & Token Budgets

Claire Prudhomme | November 12, 2026

This AI Answers covers blocked frontier models, AI agent governance, forward-deployed engineers, and AI-enabled agencies.

[The AI Show Episode 250]: Midterm Voters Reject Data Centers, AI Speeds Up Anthropic’s Research, Celebrating 250 & Nvidia’s Spy Probe

Claire Prudhomme | November 10, 2026

Paul Roetzer and Mike Kaput unpack the data center backlash, METR's Anthropic report, and the Jensen Huang probe.