10 Observations About the Top 100 AI Apps

Today on the AI Daily Brief, 10 observations about the top 100 generative AI consumer apps before that in the headlines, Amazon CEO Andy Jassy says that the company has saved the equivalent of 4,500 developer years thanks to AI. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show. Notes welcome back to the AI Daily Brief Headlines edition. All the daily AI news you need in around five minutes. One of the questions that gets asked frequently in and around discussions of generative AI is does it actually live up to the hype? Obviously we spend a lot of time exploring that here, but sometimes you just have to listen to the companies that are implementing it to get a sense of just how big a change this is making. Take for example this post from Amazon CEO Andy Jassy from a couple of days ago. Jassy writes, One of the most tedious but critical tasks for software development teams is updating foundational software. It's not new feature work and it doesn't feel like you're moving the experience forward. As a result, this work is either dreaded or put off from more exciting work, or both. Amazon Q, our Genai Assistant for Software Development, is trying to bring some light to this heaviness. We have a new code transformation capability, and here's what we found when we integrated it into our internal systems and applied it to our needed Java upgrades. The average time to upgrade an application to Java 17 plummeted from what's typically 50 developer days to just a few hours. We estimated this saved us the equivalent of 4,500 developer years of work. Yes, that number is crazy, but real. In under six months we've been able to upgrade more than 50% of our production Java systems to modernized Java versions at a fraction of the usual time and effort. And our developers shipped 79% of the auto generated code reviews without any additional changes. The benefits go beyond how much effort we've saved developers. The upgrades have enhanced security and reduced infrastructure costs, providing an estimated 260 million in annualized efficiency gains. This is a great example of how large scale enterprises can gain significant efficiencies in foundational software hygiene by leveraging Amazon Q. It's been a game changer for us, and not only do our Amazon teams plan to use this transformation capability more, but our Q team plans to add more transformations for developers to leverage. Now, of course, this is the CEO of Amazon talking about Amazon software, so you have to have that big grain of salt. However, this one stat, a piece of work that typically takes 50 developer days to just a few hours, is pretty massive and hard to ignore if it's anywhere in that ballpark. I also think that this gets at the question of what's likely to happen in the future. Are we going to see companies like Amazon massively decreasing the number of developers they have because their time gets so much more efficient? Or do they just move faster and do even newer and more complex things? My guess is that the companies that win will not be the ones who think about AI primarily as a cost saving technology, but instead those who use it to race out ahead and do new amazing things. Speaking of speeding things up, Grok 2 has apparently gotten a speed increase. Grok 2 is of course the latest version of Xai's AI software that is accessible through X Formerly Twitter After a couple days of Grok 2 being live, Xai developer Igor Pobushkin had this interesting tweet or X post if you will. They wrote Grok 2 mini is now 2x faster than it was yesterday. In the last three days two developers rewrote our inference stack from scratch using Sglang. This has allowed us to serve the big Grok 2 model which requires multi host inference at a reasonable speed. Both models didn't just get faster, but also slightly more accurate. Now I think more than the specifics, one of the things that this indicates to me is how XAI views the competition and where it can get an edge. In addition to its integration with X Twitter, the XAI team is clearly treating it as a build in public sort of project where they take on the mantle of the scrappy upstarts relative to the established giants. We'll see how that message resonates over time. Speaking of giants, one of the stories that we've been following over the summer is Google's sort of not really, but sort of acquisition of Character AI. One of the big questions was what would make Google pay such a huge premium for the company's IP as well as a handful of its team. One of the big theories was that it was a very, very expensive way to reacquire the founders of the company who had previously left Google. The latest evidence suggests that that was at least a part of it. Noam Shazir, the former CEO of Character AI, has announced as the new co technical lead on Google's Gemini, which which is similar to Microsoft soft acquiring Inflection and installing Mustafa Suleyman to run their internal AI program at Microsoft as well. If you want a sense of just how valuable top Talent is to these labs. Look no further than this deal, which, as the Information calls it, is effectively a $2.7 billion talent acquisition. The Information also had a funny post today about how ex Googlers seem to be finding that it's quite difficult to build a startup. The specific example that they point to is that AI startup H, previously known as Holistic, that raised $220 million thanks in large part to the pedigree of its four of whom are previously from Google DeepMind. Well, that company on Friday announced that three of their five founders were departing from the company. All three of those Carl Tools, Dan Wierstrah and Julian Perelot were all previous Googlers. One of them said that the founders had operational and business disagreements, but didn't elaborate further and said it was a very difficult decision for all involved. A post on LinkedIn acknowledged the transition, but didn't give any more details. In more positive news From Google, Google AI Studio product lead as well as former OpenAI staffer Logan Kilpatrick announced at the end of last week that the company had shipped a new native prompt gallery in Google AI Studio. This is a system of course to make prompting better, and it's part of a broader shift where human prompt engineering is paired with AI models themselves abilities to help create better prompts One other piece of news from the big Frontier Labs or really more of a rumor I guess. Reports started flying around over the weekend that Mistral AI is working on web search functionality. Now, this would not be that surprising as this is commonplace across comparable models, but still represents an upgrade for this set of models which is very popular with developers now. Speaking of popular and closing on the latest piece of evidence around bubble or not, over the weekend TechCrunch wrote a piece called VCs are so eager for AI startups, they're buying into each other's SPVs at high prices. Here's how TechCrunch describes it. The secondary market is where existing shareholders, such as startup employers or VCs who bought shares directly from a startup in a fundraising round can sell some of their shares to others. But because private companies like startups have a say in who can own their shares, many VCs are locked out. The VCs that do have access are setting up SPVs and selling access to their shares to other VCs or investors of their choosing, such as high net worth individuals who are accredited investors. However, as TechCrunch points out, this is different than actually buying shares of the startup stock. It's buying shares of the special purpose vehicle that controls a certain number of shares. And apparently these special purpose vehicles are selling at a premium to the underlying. One example is that an SVP that holds shares of Anthropic or XAI are being marked up to prices 30% higher than what the shares sold for in the last round. Now of course that could represent a bubble, or it could represent the market having a chance to reprice something given new information since the last round was closed. But of course that is quite a debate. For now though, if you take nothing else away, it seems that there is still lots and lots of demand for top AI companies from venture capitalists. So make of that what you will. That, however, is going to do it for today's AI Daily Brief Headlines Edition Next up, the main episode. Today's episode is brought to you by Venice. Venice is a private, uncensored, generative AI app. It accesses open source models to enable text, image and code generation without the fear of being spied on or having your data exploited. Discuss anything with Venice without concern about it being monitored, sold or given to advertisers and governments. Venice is different because your conversations and creations are kept securely within the browser, never stored or accessible by Venice. Unlike other AI apps, Venice won't tell you what's okay to say or not. Venice won't patronize you. It simply provides direct access to machine intelligence. No topics or off limits, no ideas are taboo. With Venice, you're in control of the AI as you should be. Pro subscriptions are available for $49 a year or $8 per month. AI Daily Brief listeners receive a 20% discount on Venice Pro. Visit Venice AI NLW and enter the discount code NLW Daily Brief. That's NLW Daily Brief. All one word. Today's episode is brought to you by Superintelligent, which is of course our platform that helps you learn how to use AI tools and perhaps even more importantly, gives you ideas on the best use cases that are actually going to help you achieve whatever it is you want to achieve. To recognize the end of summer and back to school, back to work. We are running our best promotion ever. When you sign up for Super Intelligent between now and the end of August using code so back your first month will be 100% free. The platform features over 600 fun, highly practical AI tutorials that get you using AI fast and with an eye to actually transforming how you get things done. We've just launched super for Teams, so if you have a group of people at your company that want to figure out how to use AI together, I highly suggest you check it out. But for those of you who are using superintelligent as an individual, once again, if you sign up for superintelligent between now and the end of the month using Code Soback, you will get your first month 100% free. Go to be super AI and check it out today. Welcome back to the AI Daily Brief. One of the things that Andreessen Horowitz does every six months or so is they catalog the top 100 generative AI consumer apps. They look at the top 50 web products by unique monthly visits and the top 50 mobile apps by monthly active users. Now that we're on our third edition, we're starting to get the ability to compare a little bit over time. And so today what we're going to do is discuss 10 observations coming out of this list. The first one, which is fairly overwhelming, is that creative uses continue to drive AI adoption. So many of these applications are about people creating images, videos, songs. There is a clear core adoption experience that these numbers seem to bear out. Second observation, there are some categories that seem to be on the come up. First of all, overall almost a quarter of these companies were new, showing that this is hardly a fixed list at this stage of development. But within both the new companies as well as the biggest movers on the list, there are some interesting themes. Music got a major upgrade this time. Udio came in on the list for the first time at number 33 that's on the web list. And Udio's big competitor Suno jumped from 36 on the web based list previously to 5 this time around. That means the only platforms ahead of Suno were Claude, Perplexity, Character AI and ChatGPT. I think that one of the things that this shows is that a big constraint on this list is how good the capabilities of different models are. Music generation just wasn't that good before Yurio and Suno's latest additions. And so now that things have gotten so much better, it's not surprising to see them moving up the ranks. However, it's not just music where we see this. The same phenomenon is also clear for video that saw both new entrants like Luma and Viggle as well as some additional big movers as well as Basically the lesson here is as the capability of a particular type of generation increases, so too does its usage by consumers. Wild, I know. Third, in our observations, perplexity continues to show momentum, and what's more, perplexity now edges out ChatGPT when it comes to visit duration. The average session on Perplexity's website is over seven minutes. Perplexity also saw more significant growth over the last quarter, although it is not the biggest grower among the top five. That prize goes to our number four observation that Anthropic's Claude is making up ground against its main competitor ChatGPT. Over the last quarter, Claude of course introduced some new features, including their Artifacts ui, which has made a huge difference in how people interact with the service. And it also introduced a model Claude 3.5 Sonnet, which for a time became more popular than GPT4O, although recently that started to switch again. In any case, from what once appeared to be a clear inevitable dominance by ChatGPT to Claude is definitely still staying in this race. Speaking of people staying in or getting into the race, TikTok parent ByteDance is spending huge effort on generative AI and it's showing up here. It had three apps debut on this list. Those include a student platform in Goth, a bot builder in Kohs, and a general AI assistant called Dubao. Dubao made both the web and the mobile list. ByteDance also has a photo and video editor Hypik, and another mobile assistant called CC, meaning that ByteDance comprises six slots across the web and mobile list. Our number six observation the one new category that shows up here might be terrifying. Andreessen Horowitz called this category aesthetics and dating. Three products from the mobile list all fell under this LooksMax, AI, Umax and Riz. I'm just going to pause here for a moment for you to start to imagine what those might mean. Given the category of Aesthetics and Dating and the names like LooksMax and Riz, pretty much they are exactly what you might expect. A16Z writes LooksMax and Umax ingest photos of the user, rate them, and give tips to become more attractive. Umax also generates pictures of what the user would look like as a 10 out of 10, while Looksmax analyzes the user's voice for attractiveness. Rizz is a tool for improving dating app messages, giving users the ability to upload a conversation screenshot and get suggestions on what to say. I will leave it to you to decide if these things represent truly the end of civilization as we know it. Number seven Observation as much as the top of these lists are dominated by the ChatGpts of the world, they also show that there's a lot of room still for more specialized applications. A huge number of these tools are effectively a single use case of something like ChatGPT or MidJourney, but turned into an entire app experience. There are very divergent theories about the long term for these types of companies. Some people think they will inevitably be eaten up by the big models, while others think that the focus on a very specific use case, especially as state of the art models become commoditized, could lead to long term product success. Of course we don't know that, but what we do know for now is that there is still a market for them at this time. Observation number eight Although this is the consumer list, there is definitely some sneaky work usage showing up here. First of all, the top of the list is dominated by tools that, while being used for consumer experiences, are also heavily used for work experience like ChatGPT, Perplexity, and Claude. Second, if you look a little bit deeper, there are lots of tools that, while not explicitly an enterprise tool, still have a ton of, if not their primary usage in the business world. Take for example Gamma AI, which is a way to build websites, presentations and documents from a single prompt. Or VEED IO, which is heavily used by social media departments to create better social video content. Now one of the interesting things at this stage is that a lot of individual users who are the ones who are bringing AI into work are also cutting their teeth on these social experiences and consumer experiences as well. Ninth observation, and this one comes directly from A16Z itself. Discord is, perhaps unsurprisingly, a leading indicator for apps that will go on to be successful. A16Z notes that in July there were 10 AI companies that ranked in the top 100 of all Discord servers by Invite traffic. That included Midjourney, Viggle, Domo, Pika, blue Willow, Firefly, OpenArt, Hydra, Kriya, and more. I'll be interested to see how this changes next time, now that Midjourney has moved entirely to the web. Lastly, tenth observation, there is still a heck of a lot of randomness in here. The thing that this reminds me of is the early days of Facebook apps or the iOS app store, when people were really just getting their feet wet and experimenting with a million different types of lightweight specific apps to see what worked. I think that if you come back a year from now, these lists will look extraordinarily different and indeed that many of these applications themselves might be doing different things by then. However, for now, this is the snapshot of what is popular at the moment and there is a lot to be learned from it. That, however, is going to do it for today's AI Daily Brief. Appreciate you listening or watching as always, and until next time, peace.