Alex Heath: Tech Journalism in the AI Era

0:04Hey, everyone, and welcome to Generative Now. This is the show where we talk to the builders who are creating the world's most exciting AI products and companies. I am Michael Magnano. I'm a partner at Lightspeed. And for this week's episode, I talked with someone else who is typically asking the questions. Tech journalist Alex Heath. Alex is the deputy editor for The Verge and author of Command Line, a weekly newsletter with the tech industry's inside conversation. He's interviewed everyone from industry titans like Zuck to founders of AI startups like Perplexity. In this episode, Alex takes us behind the scenes of his tech stories, like when he was covering OpenAI breaking news while at his in-laws on Thanksgiving when Sam Altman was briefly fired from the company.

0:47Alex and I talk about what it's like for his stories to break into the mainstream, how he develops his sources inside tech, and how the journalism industry is adapting to AI. So check out this conversation with Alex Heath. Thanks for doing this. I've been looking forward to this. We've had a lot of founders on here, engineers, designers. We've had some other VCs. We have not talked to any other journalists. So I am very, very interested to get this side of the perspective of everything that's going on in AI. You and your fellow journalists have had quite a year. Yeah, it's been a crazy year. The news cycle has been nuts and just covering this stuff and especially AI.

1:31It's unlike any time. I mean, I've been doing this for over a decade. It's unlike any time I've witnessed covering it in terms of the energy and excitement right now. Yeah, the pace alone, like just, it must be insane. I mean, for builders, you know, watching that pace force constant pivots or sort of reframing of or like resizing of opportunity, I know for sure that's disorienting. Like, how does that manifest over on the journalism side when, you know, there's a there's a new scoop hitting like every single day? You can't like develop sources over weeks or months. It's like you probably got to do within hours.

2:11What does that look like? Yeah, stuff is hitting so fast. I mean, AI alone is just feels like dozens of announcements every week that are interesting or have, you know, significance to the industry. and I think the challenge is parsing through a lot of the hot air and a lot of the stuff that is maybe just trying to take advantage of this kind of rise and then what is real like what is who are the people actually building like lasting important companies versus people that are just trying to ride the wave right and I kind of saw this with the crypto the last crypto boom a little bit with the metaverse stuff and the pandemic.

2:56This time, though, it's different. The level of interest and the fact that people are actually using these products, and a lot of people are using them and getting value out of them already, and we're only a year and a half maybe max into ChatGPT, right? It just feels different in that way. And I don't know when the hype train dies down, but it seems like it's just up and to the right right now. How do you figure out what clears the bar, right? Funding story versus product launch versus new hire? Well, I think the media in general is moving away from funding, at least I hope so, funding in and of itself being a reason to be covered.

3:41I give this advice all the time to founders who I either know personally or just I'm talking to through work. Like you have to have a narrative and a perspective that is unique and a product that stands out. And so we really look for all of that. And I look for that personally. You know, a lot of my coverage and what we do at The Verge is big tech, right? So we're covering, you know, meta, buying all the H100s or Apple's AI struggles or the morale crisis inside Google over the Gemini diversity. disaster. But I'm also interested in the early stage. I'll write about a series A company if the founders have a proven track record, the investors also are the same, and it's a really interesting space.

4:34Yeah, there's just a lot of interesting consumer applications of this stuff relatively quickly, which it feels unique, especially, again, relative to the crypto boom. You know, you saw a lot of hype, but not a lot of actual in consumer application. For big stories, right? The open AI drama, inflection, Reddit IPO, like those are known and you know sort of what you're running at. With the type of thing that you just mentioned where maybe there's a Series A or Series B company that just for whatever reason, you know, gets your attention and your interests. What is the process for that? I often say that a lot of what I do and what a VC does is very similar in terms of like the difference is the outcome, right?

5:18You're looking to write a check. I'm looking to maybe cover or build a relationship over time. And I mean, I'm in it for the long game. I'm, you know, taking coffee with founders. And even if there's something that I'm not going to cover, right? Like if I'm not necessarily going to cover the A rounds, you know, I think maybe this founder has a shot at building a really big lasting company. and I want to be there as they progress. And, you know, I do, the advice that I'm starting to give more is that founders should really lean into social and their own channels, especially when they're early in the company's trajectory.

5:51I think press, like myself, starts to make more sense when you're kind of at the D and beyond stage. But I, you know, I want to just know these people over time and I've had the benefit of being able to see that bear out over the last decade or so. And it's interesting, you see the people that work in these big tech companies too, they leave and they go start companies. And yeah, I don't know. There's not like a formula. It's more of like, how do you feel about the founder and the team? And then also just the timing of the space and what I'm covering. But I'm generally interested in anything kind of platform shift related touches consumer kind of big secular uh tech trends uh so obviously like ai's at the forefront of that right now so sticking with the analogy to vc uh you know journalism kind of like kind of like vc you know your your coffee chats your networking kind of similar to vc networking vc coffee chats and then every once in a while in vc there's some really hot company that's raising around and everyone drops everything they're doing yeah man uh when When we heard that Sam was, Altman was getting fired, that was like a drop everything, you know, start hitting the phones.

7:10Took up my weekend, took up my Thanksgiving weekend. I remember taking calls at like 3 a.m. in my in-laws dining room when we were out like traveling for the holidays. And so like in an open AI, Sam Altman getting fired moment or inflection moment, what are you doing in that moment? Obviously, you're writing, you're publishing, you're editing. But what are you trying to find out? Is there a specific person you're trying to get to? Is there a specific angle you're trying to uncover? Take us behind the scenes of the story in a moment like this. Yeah, you're trying to figure out what everyone wants to know.

7:55And as a journalist, you have this really cool position where you just kind of have carte blanche. Not always, because oftentimes the things people want to know, the companies or the primary subjects don't want the world to know, right? So that's the hard part of the job. but you are literally, you know, you're, my job is to find things out that the world wants to know. And so at the time it's like, why was he fired? Right. And we still don't really know, uh, fully, which I think is wild. And I think is something that, uh, open eye will have to reckon with at some point. But, uh, yeah, like it's finding out like, uh, the personality drama, the, like what call was made when, what is going to happen next, right?

8:37It's getting the next tick on the storyline. And we did that during that weekend. Myself and our editor-in-chief, Neil Patel, I'm really proud of the coverage we did. We broke a lot of stories during that saga. But yeah, it's what we were talking about before, building up relationships over years of people that you now know are in certain positions to know things. And that's when you pick up the phone and hey, what the hell is going on here? So I've had many moments like that in my career. I remember another fun one is when Facebook rebranded to Meta. I got a random call from a source who knew about it.

9:20I was like, hey, have you heard about this rebrand? And I thought the source was talking about some smaller thing, like even rebranding one of the apps or something. And they were like, no, they're going to change the company name. I was like, what? So that was like, That was another 24-hour crazy cycle of getting that confirmed and scooping that. And then it was a joke on SNL the following weekend. And I don't know. It's cool to see when stuff breaks through like that from the nerdy kind of Silicon Valley world that you and I are in all the time to mainstream. And I would say that, the Elon Twitter thing, and then this recent OpenAI thing, I was actually really surprised how that broke through into the mainstream.

9:58I mean, just the request to go on like mainstream TV and my parents knew about it. Altman really became a much larger public figure than I more quickly than I had expected. And I think that moment really showed that. What are the stories you're most recently fascinated by? What's dominating your mindshare right now? There's some big picture things. So how open source is increasingly rising with the LLMs, you know, just, you know, the week we're recording Databricks just put out this new model that supposedly beats Lama and Mistral on a lot of key benchmarks. And they trained it for$10 million.

10:39And so the rapid pace of innovation and open source and costs coming down, what that means for the highly valued closed private model companies, I think is going to be a really interesting, messy storyline for the next 12 to 18 months. and more company specific. I remain fascinated by Google and how they will or won't rise to this moment. And can they overcome the missteps they've already had with Gemini? How big of a market is this going to be relative to search? Does it eat search? Does it complement it? I actually think the jury's still out. You've got really interesting startups in the kind of nexus of search and AI.

11:24I just interviewed the CEO of Perplexity in my newsletter command line. And, you know, I'm always trying to meet interesting and interview interesting kind of people at the forefront of using AI in different ways. Another thing is just the middle. And I think another thing is where consumer hardware and AI meet. and whatever OpenAI ends up doing with Johnny Ive, for example, or a lot of these really interesting consumer devices that use LLMs, Rabbit, Humane's not using LLMs really, but I'd put them in that category, these wearable companies. I have the meta Ray-Bans with the multimodal AI enabled on them.

12:13And it doesn't take long to realize like, Like, oh, this is maybe like what everyone has actually been talking about when they're talking about what comes after the smartphone. And that feels closer. You know, I've been covering augmented and virtual reality for a really long time. And it feels closer than ever now because of how contextual AI actually makes sense in this kind of egocentric first person modality. and I don't know, I think we're in for a wild ride of consumer hardware kind of innovating on LLMs in the next couple of years. Do you agree with that? I totally agree. I think you and I touched on this a little bit at Upfront a few weeks ago.

12:56I think obviously there are potentially really new and cool hardware devices to be built, but I also think existing hardware devices now with the benefit of having an LLM attached to them can do some really, really interesting things. And so, you know, this is probably the most obvious answer in the world. But, you know, when my Apple Watch gets an LLM in it, powering Siri, that is going to be an absolutely life changing experience. Right. I am so trained to talk to Siri, but obviously Siri does not do nearly as much as I would like it to do. And if it could do everything an LLM does right on my wrist, that's a that's a really, really powerful shift.

13:33So I totally agree. I want to touch on Google a little bit. You know, you mentioned the Gemini fallout. That was very, very recently. What are you, you know, through your reporting, which I'm sure you're all still digging in on this. What are you hearing from inside the company? Not necessarily about what their plans are for the product or how they're going to bounce back, but like more of the sentiment. Like, what do you know about how the tone is inside of those buildings for what comes next? I think it depends on the team. I mean, Google is such a sprawling conglomerate. The research team, the team building the actual AI, they're pretty jazzed.

14:13I think Gemini as a model is impressive, the context window aside even. And they're psyched about the work. They have the resources. They have the buy-in from Sundar and Sergey Brin is back in the building, giving that kind of founder weight to the project. at the same time I think they're suffering from a lot of this just kind of organizational fiefdom that has been real in Google for a very long time but is now becoming much more of a problem when they're trying to move a lot faster you know I always joked over the years that like I wasn't the only one to say this but Google is very much like Game of Thrones where you have like the different houses and different corners of Westeros who all think they answer to a different person.

14:59In the last 12 months, the big project has been how do we unify these very different teams? So it's like merging the research teams, brain and deep mind, which is still kind of ongoing, but also just like the product team and the search team. How do you get them to work better with the research team? And how do you avoid things like what happened with diversity stuff, which I think and what I've heard through reporting could have been caught if teams had been more aligned kind of in the initial build up to that release. Yeah. A lot of it is like knocking down heads and knocking down walls and getting collaboration going.

15:39And this is where startups, of course, have the advantage of not only moving faster, but just being smaller and not having this massive kind of not just it's like organizational debt, not technical debt, but just this, you know, the org chart problem. And so Google is like suffering from being such a huge thing that now needs to move fast. And I think you could contrast that with Meta, another company that I follow very closely and talk to Zuckerberg pretty often. There you've got a founder still who is very dialed in and they have less. I'm not saying they don't have it. They're huge still. They definitely have the organizational problems, but I would say they have less of what Google is dealing with.

16:27And as a result, they're shipping, I think, a little faster on a lot of this stuff. So it's really, it's, you know, I've never seen a wave like this where not only does it build up so much energy and excitement in startups, But, you know, big tech is like completely re-rotating around AI at the same time in a way that I just didn't see with crypto, with metaverse, et cetera. So I think the jury's out on like where the value accrues over time. And if you view this as a platform shift, you know, history would tell you that it will be startups that end up, you know, accruing the most value over time that usually the incumbents have a hard time with shifts like this.

17:13And that's the getting back to Google. That's Google's dilemma. It's like, can they can they shift fast enough and not compromise their their business to really answer to this moment? Do you think Google internally is worried about open AI? Open AI seems like the most direct, like the most obvious direct threat or, you know, some combination of open AI, Microsoft. But do you think they think at all about companies like perplexity? I think they're worried about anyone carving off any percentage of search. Perplexity is obviously very small, but things always, you know, big things start small. They have to at some point, right?

17:49Right. Yes, obviously, OpenAI has been the technical leader in just producing the best models. I think what's been really interesting over the last month or so is the gap between OpenAI and everyone else is quickly shrinking, right? And so OpenAI is talking about GPT-5 later this year. If that's another massive leap, that really needs to put them 12 to 18 months out from everyone else for them to maintain their lead just because of how quickly these models are commodifying. I think OpenAI feels that pressure. They also feel the fact that Microsoft is intentionally diversifying away from them with hiring the inflection team, the Mistral deal, etc.

18:39At the end of the day, you need distribution. Good tech is great and it matters, but to really be a massive consumer business distribution wins. I mean, we saw this Google having the foresight to pay Apple what is now on the order of billions a year to be the search default on iPhone. Apple being the richest company in the world because it owns the most important mobile hardware platform in the world. OpenAI doesn't have hardware. They don't control the endpoints to the consumer. And it's why Zuckerberg and Meta is spending DARPA level money on building what he thinks may be the next hardware platform.

19:19He doesn't want to be a surf on Google and Apple's mobile platforms. And so that's the real question for OpenAI, Anthropic, a lot of these model companies is just investing in models and building the best models enough to win here? Do you actually, you know, if you're meta and you have 3 billion users and you can do something that's even half as good, but it reaches 3 billion people, what does that do to the company that's technically has a better model, but doesn't have even, you know, a 10th of the distribution? We were in the kind of like initial, this is so cool. This tech is amazing. The world is changing mode for, I think the last like 18 months.

19:59And I think we're entering the, okay, got to figure out business model, got to figure out distribution phase of this now. Otherwise, you know, there's a lot of money that's been plowed into this sector. And, you know, it has to, this has to be figured out for there to be returns. Right. We can't just like keep riding on like the cool tech and like feeling of like, this is the future for, for much longer. I want to touch really quickly on the inflection Microsoft story, which you covered, you know, I think. I forget the exact words you used, but it seems clear that this was some sort of possible workaround to regulators and the lack of ability to do any real M &A right now.

20:48What are you hearing in your reporting about how this eventually plays out? I mean, do these deals become the norm? Do these deals then also get squashed by the regulators? Does this force the regulators to rethink how they're regulating M &A currently? Like, what are the second and third order effects of what just happened? Big tech has the red light on big acquisitions, especially in something like AI. You know, there's already a Biden executive order. The government is really concerned about concentration in AI, as they probably should be. But it's early and the level of like interest from the government is different than any other kind of tech wave.

21:26Usually it comes in much later. So big tech knows that they can't buy in this space, much less really much of anything. I mean, you just saw Figma getting blocked, that acquisition by Adobe. And so, yeah, what you're seeing is like, I thought the inflection thing was pretty brazen on Microsoft's part because they tried to hide it, but they're hiring basically the entire company, paying a licensing fee that is conveniently just enough to make the investors in inflection whole, a licensing fee for a company that is now running on autopilot without its team. And so it's like, you're not technically acquiring the equity or the IP, you're just making it worthless and taking the team.

22:14And you're going to see a lot more like this because, again, because of the rise of open source and how quickly the cost curve is coming down on building AI, I think you're going to see a lot of these AI model companies go under because really it's going to end up being like, I think one or two, I think it's going to be similar like mobile. There's only going to be one or two lasting independent players in this space that aren't already tied to a hyperscaler or a mega cap, you know, Magnificent Seven company. And, you know, if inflection couldn't figure it out with co-founder of DeepMind, Reid Hoffman.

22:53What does that say about everyone else? It's just it's a tough slog. And so, yeah, everyone in big tech is going to be doing more of these kind of aqua hires, these these fake acquisitions because they do an acquisition that gets put in FTC red tape for 18 months, potentially unwound at the end. And no one wants that with how quick the industry is moving. You got to you got to move fast here. So it was smart of Microsoft to do what they did because they get the acquisition without actually making an acquisition. It's a matter of whether this will catch up. I don't think regulators can come in and say like, oh, we're going to sue you because you did a fake acquisition.

23:33I think Microsoft knew what it was doing here. But over time, it will probably mean more stringent regulation on these kinds of deals. And I would expect that. I wouldn't expect that in the near term. That's not like a this year thing. But yeah, you're going to see a lot more of these kind of over frankly overvalued uh no revenue ai startups looking for lifeboats like this um because uh it's it's tough out there in the in the ocean right now let's talk about the future of journalism and ai maybe maybe starting at the individual level aka you uh as a journalist how are you currently using ai in your work or are you i am i would say cautiously and slowly um Yeah.

24:20Hallucination is a huge problem, right? Yes. And it's something that is a problem for me when I deal in facts and needing numbers to be correct and references to be correct. So I would love to use LLMs more for the padding of stories and all that, but I just don't trust it. I actually had this conversation with the CEO Perplexity. Like he thinks they're solving it better than most, but, and I've, I've talked to Sam Altman about it and he thinks like, we won't be talking about this problem in a few years, which, you know, that would be awesome. But until that's solved, what I really like is the more creative applications.

25:04So, you know, I've got, you know, notion Q and A, which has been kind of helpful for summarizing and kind of engaging more creatively and in a brainstorm way with my notes. A friend of mine has this interesting product called Lex, which is basically like an AI assisted Google Docs. And Google is doing some of this in work in their software natively to which I've tried, but I'm not really using the image generation stuff. I don't haven't really, you know, we have an amazing art team at the verge, you know, we like, we are kind of new thing internally is like, if you're writing or your art could be done.

25:44mostly by ai like uh what are we doing right it's like yeah why are you working what were the value we're adding is like our humanity uh and right right uh and so i actually tried to intentionally steer clear of of overusing it um but you know i have you know open ai's uh audio thing as my action button on my iphone um oh wow and i've been trying that as a serial replacement and And when it works, it's awesome. And when it doesn't, like when it makes up where the moon is or something, you're like, oh, okay, this is why this is not ready to be like default in the phone or an actual Google search replacement.

26:27And again, the multimodal in the Ray-Bans is a really fun thing that I've been trying where you can literally just like look at a plant and say like, how much water does this need or something like that. I'm going on a trip to Japan in a couple months and obviously don't know the language. So I'm actually very interested in using them there to be able to look at things as I'm walking around and translate them in real time. So that's super exciting, like wearing this stuff, especially when the application, you don't need to necessarily have like a factual search answer. It is really exciting. But yeah, I don't know.

27:05It's different for me in that way because I covered the crypto, last crypto wave too. And it just felt like a lot more of fluff. And I got it conceptually and I still do. And I still think a lot of these L1s and L2s are really interesting and have real world applications. But I don't know, man. It's different this time. This is stuff you can actually use. Totally. And that's exciting. Yeah. I could see the hallucination thing totally makes sense. Like you, you can't be having chat GPT, right. You know, write your piece on that point though, like getting to what I was saying earlier, I cannot wait for a model, whether it runs locally or not that.

27:48And just all of my writing in a privacy safe way where I know it's not being like sucked up into some bigger training cluster and shared with everyone else. Cause I have notes that are like very sensitive that like, I haven't like anonymized that could get people fired. Maybe Apple do that. Yeah, something like that where my notes app is an LLM of all of my writing. And I can say, here's my... Because there are some reporters who love the writing process and I'm just not one of them. I actually just love reporting. I love talking to people and learning new things and connecting dots. So if I could just dump all that into a doc and say, here are my notes for the newsletter this week, write this in my voice.

28:24I would pay$50 a month for that. That's cool. Because it saves me time. And that's the promise of AI, I think, is that it can accelerate and solve parts of what we do day to day that is either just drudgery or things we don't look forward to. And for me, that's like a huge thing. Like when I finally have that in a product that I trust, I will use that all the time. Yeah. One of the things I think about with this right now is the tools will get better, right? Like the product you just described, like that seems like it's going to exist, right? Yeah. And there'll be lots of other products that'll kind of give you as a journalist superpowers.