Edit labels for this show, save them in this browser, or download a JSON override for the production folder.
Transcript Playback
2025 And All That: The Path to Abundance
Human Transcript
Timed transcript
Blocks are grouped by speaker for readability. Expand a block to inspect word-level timing.
Speaker 1
Hello, everybody. It is late in December 2025. I'm going to do one more final. That was the week's show. We're going to look back at what happened on the tech front in 2025 and look forward to 2026 with my old friend Keith Teer, the publisher of That was the week newsletter. Keith, let's look back at 2025. Your friends at Google LLM put together a nice slide of what happened in 2025, you're calling it, or they are calling it at least, the Year of Intelligence and the Great Compression. Let's deal first of all with the Year of Intelligence. What does that mean?
Well, you know, there was this huge doubt a year ago about whether LLMs could scale, whether they were really, you know, able to contribute to human experience, as it were.
And I think now the... Can I just ask one other? Can we ban this word human? Because it's such a meaningless word. There's no such... Certainly you and I are not able, I don't think, to talk about any other experience except... it being human because it's just by definition, experience is human.
No, look, AI clearly is a huge contributor. to anything you choose to use it for now. And that's beyond a shadow of a doubt. I've noticed even Gary Marcus has stepped back.
Even Gary Marcus, who actually comes up all the time, whenever anyone mentions skeptics, it's always Gary first. And I'll try and get Gary on the show before the end of the year to actually give his conclusion. I think that'll be different from yours, but go on.
Well, he has a true north, which is LLM's don't know anything. And he's not wrong, by the way. If you're purely scientific about it, LLMs are not intelligent, as it were, because they're not consciously thinking about problems and solving them. But they're remarkably good, nonetheless. And they contribute in many, many ways, depending on who you are and how you use them. And I think that's now incontrovertible. So we passed the point of skepticism. Now it's, will the bubble burst? We've moved on to the economic thing. Okay, it's great, but will the bubble burst? I think we're almost to the point where nobody thinks that. Almost none of the predictions for 2026 include the bubble in course bursting. So I think we're almost beyond that.
So now it's, well, how... Well, let's talk about an idea. Before we get to 2026, let's talk about 2025. The bubble clearly did, if it is indeed a bubble, which I don't think you even think it is a bubble, but it certainly didn't burst in 2025.
Well, it means you can't live without it, just like you can't live without water or electricity. You know, you could try, but it wouldn't be very pleasant. AI is becoming ever present. in most professionals' lives and many individuals' lives. It's a utility in the sense that you start to take it for granted, it's assumed, and you know you'd be worse off without it.
Or the internet. I'm not convinced about... I think for those people who use it, it becomes a utility. But what's the proportion of you? I don't want to use this word humans. What proportion of us, Keith, use it?
We always have this. Every time there's a new technology, people always come up with this. It just means that any time there's a new technology now, it's faster. That's just the nature of things. Yeah. Okay, so it might be. Shall we say intelligence is becoming a utility? I'm not sure it becomes. There's a big difference. It's on the way to become. Yeah,
Okay, and then we have the Hollywood ending. So we've got intelligence becoming a utility, the browser becoming an agent. We've talked about this. What exactly does that mean?
It means it can take actions. For example, it can manipulate a spreadsheet or a PowerPoint and so on and so forth. So you can load a web page, any web page, and you can, if you're in Chrome, you can now use Gemini. If you're in the ChatGPT browser, you can use ChatGPT if you're in the Perplexity browser, so on and so forth. They've all got browsers now. And the agent embedded in the browser can see the page you're on. and you can ask it to do things or explain things within a page. And in that sense, the browser is turning from being a screen, to being an interactive.
Right, so it's turning. It hasn't turned. And I take your point, but when I use Chrome, for example, the top of the page is Gemini and the bottom of the page are still links. So we're in the process. No disagreement there. I hate this word, pivots, almost as much as I hate the word human. Policy pivots from panic, all the Ps, to opportunity. What does that mean?
Well, a year ago, all the dialogue around AI was about safety. Now it's all about clearing away the roadblocks or the friction that slows it down. The administration's decision this week to do a single federal rule about authorizing AI.
When you're talking about this pivot, the P word, You're talking about official Washington DC. You're talking about the shift from Biden to Trump. I mean, there's still a lot of people who are very panicky on AI. They're just not in power. Yeah, I think... Using another P word, policy, pivot, panic, power.
Yeah, honestly, I'm not even sure Trump was there initially. I think you can credit David Sachs with quite a lot of influence on shaping policy. And his shaping of policy doesn't remove safety, but it subordinates it to progress or moving fast, let's say.
Generally, I think it's a smart thing. I think technology doesn't wait for government. And if government's too slow to get behind it, it will get ignored anyway.
So let's move on. Policy pivots from panic to opportunity. I agree if we're looking at DC, although I think there are a lot of people who are more and more, for better or worse, more and more fearful of AI. This one's an interesting one. Power and capital consolidate. I think that may be right, especially when you're talking about Trump and guys like Sachs and Andreessen. What does that mean?
Well, the details would be a long list, but to sum up the details, more and more money is coming from fewer and fewer sources into fewer and fewer companies. And so there's a small number of horses in the race. They're all raising large amounts and mainly from the same suspects. Today, OpenAI is seems to have leaked that it's trying to raise $100 billion on a valuation of just under a trillion, 820 billion, I think it is. And that's symptomatic of the trend. They're certainly not the only one.
And you're not someone who's particularly thrilled about this as someone who represents the startup, more of a traditional startup guy. Is that fair? This is not a development that you're thrilled by.
I would say it is smart money chasing big wins. It is smart money. It's not like retail investors are doing this. Retail investors are investing in secondary markets. I thought you liked retail investors. The ordinary man on the street, Keith. Aren't they the smart ones? Well, that was my point. My point is this is not driven by retail investors because most of these companies are private, except in the secondary markets where it is to some extent. This is driven by Fidelity and various Middle Eastern sovereign wealth funds, large allocators who are trying to turn something into a 3X in a short period of time and probably will succeed. My point about startups would be this. A lot of startups are getting funded in the AI space as well. And they will end up being large winners in the application layer above AI. There are many of them. But everyone else that isn't in AI is being starved of funds.
And I think people can hear some banging in the background. That's the ordinary man, Keith, in Palo Alto trying to get into the deal, isn't it? It is. It is. Maybe I should open the door to stop the banging. Just give me a sec. Keith is about to open the door and we're going to see the appearance of the retail investor, the ordinary man. There you go. This one, Keith, the fifth point in your summary or in LLM summary, I agree. I couldn't agree more. Institutions are outpaced by change. I mean, it's astonishing how paralyzed institutions or many institutions, educational, old school, VC, political, seem to be in this age of just such remarkably fast and dramatic change.
Yeah, it's really where established process and employee bases meet the future and can't adapt to it. It's super hard to adapt when you've got sunk cost in buildings, in using education, the example, in curriculum, in testing, and at the same time, teachers, teachers' unions. It's super hard to change.
Yeah, the old world is in dramatic crisis, and depending... I mean, the left and the right are dealing with this threat in very different ways. But this is the essence of our current politics on the left and the right. And the disappearance of the center is this outpacing of institutions to make them essentially irrelevant. And yet they're not irrelevant because they still exist.
Yeah. I mean, all it really takes is for an individual with authority to ask a simple question. How can we use this to benefit, in the case of education, the students? And there'd be lots of good answers to that. And I think there are some people probably asking that and trying to get things done, but it's so little and so slow.
Too little, too late. I think I mentioned this story before. A few weeks ago I was in DC, a reunion of some old college friends. One of the guys there was a professor at Berkeley and he was telling this story and he was horrified about how students are refusing to read the text because there's no point anymore. They can use an AI to summarize. And of course he was saying, this is absurd, but actually, The real challenge is getting him to realize that there are other ways to become smart rather than reading a text. And if you've got an AI that allows you to do it, then you've got to think of other learning processes.
Yeah, exactly. I mean, you know, I remember reading Sons and Lovers when I was a young... Illegally. Wasn't that bad when you were a young man, Keith? No, D.H. Lawrence was de rigueur, as they say. I thought you grew up in the 50s. 60s, but close. Anyways, I remember reading it and taking the longest time to understand the themes, probably not even abstractly separating the themes from the story, and then having to do exams, you know, they're a little bit more abstract, when I hadn't really digested it. And if I could have understood the themes and digested them and thought about them, I may have had some original thoughts about it, but didn't. So I think AI can basically help rather than hinder.
Were you the sexual god when you were young, Keith? Oh, easily, definitely. You were. Keith Teer, for those who aren't able to watch this, looks as if he stepped right out of a D.H. Lawrence novel, Sons and Lovers. Finally... this was a bit of a throwaway keith because this is again you wearing your anti-doomer hat the real risk isn't extinction it's consolidation is that keith tier speaking is that llm speaking because there's still people who believe in
extinction it's a little bit llm speak um i you know if you ask me is there any risk with ai i think i'm on the extreme side of saying no not Not a little bit, but none. It's mainly all opportunity and reward.
Yeah, I'm framing risk a bit differently. If you take teachers as a collective, I think there's only opportunity. And I think if you take... You know, any job, really, there's only opportunity. But for the individual, that can mean there's no job for them, for sure. But that doesn't mean there's no life for them.
Well, we're not going to get into it. That's another subject. It may be another life in some ideal world, but in 2025 or 2026, it's not much of a life. Let's get to some highlights. Who's your team of the year, the company startup tech company of the year for 2025? Well, let's pick one of those labels. And you're not allowed Manchester United.
Well, I'll come with someone else, but just to stand back a second, if you think of SpaceX, Grok, what's happened with Tesla... his brief excursion with Doge. He's had a huge impact this year.
Well, that's fair. I accept that if you acknowledge like the Time Magazine person of the year that I think they once made Hitler person of the year, I'm sure they've made Trump person of the year. It doesn't mean they do.
If you want me to ask, so it's a favorite person of the year. My favorite person of the year. Hmm. My favorite person of the year. I think I'm going to say Sergey Brin.
Yeah, everybody loves Sergey, don't they? I mean, you can't hate him, although he got swept up in the Epstein thing. Well, the thing about Sergey is he obviously is super wealthy. He's good looking, he's super wealthy, he's progressive.
No, but the time he went in, DeepMind was a side project and the Google AI was run by other people. He went in and he insisted on some very fast outcomes that happened, one of which was that Demis Hassabis was put in charge of overall of Google AI. And I think he's...
he wasn't in charge overall of everything i think sergey could be our we might make him our that was the week keen on person of the year for silicon valley because He's smart. He's personable. He's good looking. He's not a... He's geeky, but he's not a total geek. And who else could it be? He doesn't say really dumb things like, in the future, there won't be any money.
Yeah, it's... It's thunder. It isn't Mark Zuckerberg. It isn't... It's not Mark Zuckerberg. It isn't Microsoft. I don't think it's Sam Altman. He's had a good year, but I don't think he's the... Has he had a good year?
I mean, you and I disagree on him. I'm willing to talk about Sam Altman in contrast to Musk. I still think that he's a bomb waiting to explode. I think he's a bit of a liar. And I think the more fibs he tells in the end, open AI, it may not blow up. I acknowledge it's a real company worth hundreds of billions of dollars, but I'm not convinced that it's going to be remembered as Sam Altman's company, but maybe I'm wrong.
Look, he's not the technical lead, but he's- No, no one's pretending he is, but- He's navigated shark-infested waters on all sides very favorably. I agree, and he's done very well with Trump, it would seem. With Trump, with Microsoft. with his former co-founders. He's kind of neutral. Even Elon Musk is somewhat neutralized.
Well, I don't know how neutral. I mean, I think to be fair, I don't care for his politics. David Sachs would probably be the person of the year in tech.
He's got to be on the list. Yeah, I absolutely agree.
Words and timings
He'sgottobeonthelist.Yeah,Iabsolutelyagree.
Speaker 1
How powerful is he though?
Words and timings
Howpowerfulishethough?
Speaker 2
I mean, he's a voice in Trump's ear and Trump seems to listen to him so i think he gets a lot of credit for the ai and crypto changes that have happened that those are huge changes for businesses yeah um the
other person we might put forward is amadai and anthropic i'm a big anthropic user i don't know that much about amadai but he has injected an element of concern and responsibility which the others haven't what what is remarkable about
the original founders of open ai of which he is one uh and ilya sutsukeva is one uh uh andre i forget his second name they all are super thoughtful opinion shapers and technically, you know, skilled in AIs.
It's almost like the... Saskeva is his name. Yeah, I mean, I think what we're going to remember about OpenAI and that original team, it's rather like the PayPal mafia. They'll come to shape... the AI world.
Yes, and including Feifei and Maria, is it the Italian, Matt Lube, who was the CTO at Oplan. There's a lot of great people that started there and are all now doing important things in their own right.
Yeah, I mean, I think, Keith, when we look back at, and I want to come to 2026 now in a second, but when we look back at 2025, it's a year of becoming. Nothing strikes me as happening of, I mean, if I said to you, what was the most important event in Silicon Valley in 2025? None really come to mind, do they?
No, I think it's a list. It's not one thing. In other words, nothing much happened. And becoming is a good word for it. It's, you know, early innings, as they say in baseball.
Yeah, as they say in Germany, Heidegger would say it's the year of becoming. So let's, in Yorkshire, they say early innings, and in Germany, they say becoming. Let's move to 2026. Is it going to be the year, Keith? Yeah. And your LLM talks about the year AI's build-out bites back. It sounds a bit Yorkshire-ish. Is this going to be the year where it actually becomes, where something happens, where everyone recognizes now that AI is the thing in it? To borrow another German term, philosophical term, the thing in itself.
I think we're going to see huge spending on infrastructure. Yeah, but we already have. I mean, that's a reality. No, we've had huge commitments, but they haven't actually built much to the completion yet. So this year they will. I think the idea of data centers in space will become tangible and real. So it won't just be Earthbound, it will be... That's a bit abstract.
I mean, in 2026, I would have thought that's more like 2036.
Words and timings
Imean,in2026,Iwouldhavethoughtthat'smorelike2036.
Speaker 2
I'd be shocked if one or two players don't launch satellites that are data centers in the next 12 months. I think they will. And also, what's the big deal about that? Well, because the sun's capacity to produce electricity in space is many, many times greater than it is on Earth. and they basically believe they can get free power.
That's a big deal. Okay, well, that's very abstract. Let's talk about some of the more concrete things about your predictions for 2026. You talk about a mini-layman moment for AI, which is interesting. Just a mini one? There may just be a layman moment. Is that the bursting of the bubble, Keith?
No, but it means that there's probably going to be a loser. It may be a debt holder who's holding, let's say, Oracle debt. It's hard to know where, but the ratio of debt to equity and revenue is getting heavier and heavier on the debt side. And at some point, somebody is going to have made a bad bet.
Well, they already have. Does it really matter, though? I mean, if someone's stupid enough to lend to Oracle, they probably deserve to lose all their money.
Well, the mini-layman, I think there will certainly be a mini-layman moment, but it's like a stroke. You have a series of mini-strokes and then a big stroke. So the issue is not a mini-layman, but a major-layman. What about the AI power crunch beginning? How will that manifest itself? What does it mean?
Well, it starts, first of all, with the fact that China doesn't have this problem. It has almost unlimited power and is building... renewable energy at a rate no one else in the world is, and therefore its data centers won't be inhibited by power. It was announced this morning that somebody in China has invented an optical chip that's 100 times more powerful than the Nvidia chip. I don't know if it's true, because I haven't gone and done my research, but it wouldn't surprise me that innovation coming out of China is gonna drive rapid progress. And the US, you know, to stay in the game is going to have to find energy. And so the energy crunch really is a way of defining the problem. The problem is we need more energy. And where's it going to come from?
And we've had this weird, almost surreal merging of Trump's media company into some nuclear fusion company. Is that an example of this? I can't make any sense of that one. I mean, so it's even bizarre for Trump world. It reminds me of, um,
Morgan Stanley Banker, back in the late 90s, when I thought about merging a company with another company, told me that if you tie one brick to another and throw them in the pool, they sink faster. I think it may be a case of that in those two cases.
Yeah. I mean, I agree with you that the AR power crunch will begin, but again, it's beginning. It won't be as sudden as if all these... sensors are going to go dark or that we can't use ai correct it won't so that's it
may be friction in the scale of growth you know sam altman always makes the point that if he could double energy he could double revenue today and he can't double energy so he can't double revenue so there's a a relationship between revenue and
energy And that's a classical, I mean, that's an excuse for not making a profit. You include, and you've talked about this a lot, AI agents starting to do real work. What is real work, Keith? What does that mean?
Work currently done by humans, probably. Well, that's already happening, isn't it? The classic, well, it's happening on the margins. So this is talking about it accelerating to the point where it's decisive. And the places you would most look for it are customer support and sales. That's where human beings engage with other human beings. That's where people interact with other people, sales and customer support. A lot of the customer support side in particular is based off of written rules that the person on the other end of the phone read from. easy to automate completely. And on the sales side, it's driven by business cases to a customer, again, that are discoverable and reproducible. So I would expect that's where you're going to see it first. I don't think you're going to see physical AI robots doing human jobs that are physical yet, except on production lines.
Yeah, he needs a lot of home help, that guy. This one, I think, is more real. I mean, these are all vague. I mean, AI has always been important with customer support. I don't see anything happening in 2026.
Well, it's about, will there be such a thing as a person doing customer support a year from now? And I guarantee you there will be. And what percentage is less than there is today?
Well, I don't know, three, four, five. But that's, again, becoming. And you don't think it's going to be 50%. I don't know what it is now. So I'm doing a show with Julia Hobsbaw on the future of work. Maybe we can ask Julia on that. And I know she's coming to the US next year. She'll do a show with the two of us. So this one, I think, is real. Political backlash to AI gains intensifies. You forwarded me a piece about Bernie Sanders wanting to put a moratorium on centers for intelligence. which astonished me. I mean, I can't believe you even said it, but I think that does suggest that this is becoming a real political issue.
Yeah, and note the words there. The backlash isn't to AI, it's to the gains of AI. So a gain for AI is often depicted as a negative for the population. And the backlash comes from a desire to support or protect people. And that's clearly where Bernie Sanders... People are better than humans, aren't they? Yeah. And that's where Sanders is coming from. It's a mistaken framing. You know, AI is...
there will be a political backlash for sure, and it will come from the left of the Democratic Party, and it will seek regulation. Big tech will continue to be their boogeyman, and it will get louder as we get closer to the November elections.
I doubt it. I think they're less motivated to be anti-AI.
Words and timings
Idoubtit.Ithinkthey'relessmotivatedtobeanti-AI.
Speaker 1
And do you think that Trump will put all his chips, so to speak, into Silicon Valley, into his friends like David Sachs, rather than the Bannons and the Marjorie Taylor Greens of the world?
That's why he merged his company with nuclear fusion. This one I think you're wrong on. Education systems are forced to adapt. They're not. They're just going to sit there. They don't change. This goes back to your previous slide on 2025. Institutions are outpaced by change. There's no institution which is more outpaced by change than educational ones, in my view. For better or worse, I don't want to celebrate that. I'm not saying it's a good thing. It's just the reality.
Yeah. Look, I think we're going to see the first AI university launched next year. And we're going to see students enrolling in it. And I think that will produce the established educational system to try to compete, probably fail, like all dinosaurs usually do fail, but they'll try. And I've seen it already in California where they're passing rules now to allow AI in schools. I'd note that yesterday, OpenAI announced that they'd sold bulk packages to 14 universities in the US for very large sums. So there is beginning to be an awareness that you have to integrate, not fight it.
And of course, we will have many articles about how real-world universities or colleges with people teachers are so much more valuable, which I think is true. Finally, your sixth point, I don't know if it's yours or Google LLM, regulation shifts from content to defaults. What does that mean?
A default is... locking in a user by having defaults that can't be changed on an application. And what this is pointing to is that there's going to be an increasing focus from the regulators on, just like they did with Windows, they made it possible to use different browsers by regulating that you had to. They're going to start regulating that you can move from, say, ChatGPT to Anthropic and take your data with you.
I mean, firstly, I'm not sure that's going to happen next year. And secondly, big deal. I mean, I think all these suggest, Keith, that really, whether it's you or me or Google LLM, none of us have any idea of what's going to happen in 2026. We know there's going to be big change. We know that AI is a technological force that is changing everything. But it's hard to actually predict anything concretely.
It's easier to predict what won't happen. You know, there won't be a bubble burst. There won't be a deterioration. Well, there may be a mini layman. Yeah, but that's different.
What else won't happen? The framing of AI as a race between nations will persist and maybe even get stronger, even though I don't believe that's the right framing. I think the America first mentality, which is becoming all pervasive even across the political spectrum, will probably get more intense.
Very realistic, I think, based on everything I'm hearing. I think for sure Anthropic is considering it seriously. I think OpenAI will... react to that and consider it. That said, there's a huge amount of money in secondary markets where these companies can go and sell shares for money, put money in the pockets of their investors and employees without going public. So it isn't a binary outcome. Either one could happen or both, but it's more likely there'll be IPOs than it was this year.
So it's not going to be a Netscape moment. I don't believe so. Finally then. I'm not sure we need to go to 2026, Keith. Maybe we should just skip to 2027.
Oh, is that official? You've just announced it. That's major news. Here's the question. If SignalRank goes public and you become a billionaire, the new billionaire on your block in Palo Alto, will you still do this show? The nature of SignalRank is that...
If I was ever to become a billionaire, it would be many years from now because we're an index that grows surely but slowly. Should I knock you up for cash for doing this? No, but I will still do the show. It's my fun of the week.
Let's end, Keith, on what product, and I know what I want. I, speaking of Christmas, we're not supposed to do Christmas in this Jewish household, but I'll do Christmas for a bit. I know what I want for Christmas in 2025, 2026. I know the tech product I want. What would you like? What's missing from your life that would just change everything?
I would like my AI to be manifested in a physical robot. Not a dog. That I can talk to like it was a human. Hey, Luna, Luna, stop it. Thank you. Say that again, man.
Yeah, I want a dog muzzle, a dog muzzle. That's what I want. No, I want my AI to take physical form. where I can talk to it and interact with it as if it was a person and still be an AI. That's what I would want. And it can do physical tasks that I can ask it to do.
Well, you've read Clara. Have you read Clara and the Sun? I have actually, yeah. So, yeah, I'm not sure I want that. um i think we will get that but i guarantee we won't get that in 2026. i'll tell you what i want i want um adobe to come out with a ai version of premiere pro where i know exactly everything i want but i don't actually have to do the programming i can tell it exactly what areas of the video I need to edit, how to integrate the audio, and for it to do it without me having to fiddle around with Premiere Pro. That's a much more realistic wish, isn't it?
They'll do clips and stuff. You know, intelligent video editing is probably not going to happen because it involves the AI knowing your human decisions about cuts. And I think that I've never seen it be successful yet. The AI makes decisions that I wouldn't make personally.
Well, you've just ruined my Christmas, Keith. We are going on an extended break. Keith's going to South Africa, then I'm going to DLD, and then Australia. So I don't think we'll be back until February.
Oh, dear. Well, you're not coming on keen on with AI. That's all I can say. You can do it on your own. That was the weak platform. Happy holidays, Keith. Safe travels in South Africa. And we will see you all again in February. Bye.