Skip to content

Podcast

The Top GenAI Apps Right Now

The AI Daily Brief: Artificial Intelligence News and Analysis

Source ↗ ← All highlights
  • Larry Page’s AI Ventures
    • Google co-founder Larry Page’s new AI startup, Dynatomics, aims to optimize product designs using large language models.
    • His previous venture, Kitty Hawk, focused on developing electric flying cars. Transcript: Nathaniel Whittemore We kick off today with some startup news, but not just any startup. Google co-founder Larry Page is, yes, pivoting to AI. According to reporting from The Information, Page is building a new company called Dynatomics. The startup, which is still in stealth, will apply AI to product manufacturing. The Information writes, Page and a small group of engineers are working on ways to use large language models to create highly optimized designs for a wide variety of objects and then Have a factory build them. Not totally clear what that means, but Dynatomics is clearly part of a new wave of companies trying to marry AI to manufacturing to produce next generation products. Another company called Orbital Materials is using AI for advanced materials research to produce technology from batteries to carbon capture devices. X is providing AI-driven simulations for testing parts of automotive and aerospace applications. Instrumental is one of many using AI Vision to improve monitoring and maintenance processes in factories. And if Page’s last project is anything to go by, this new startup will be even more ambitious. From 2016 until 2022, Page was working on electric flying cars through his company called Kitty Hawk. The startup built over 100 vehicles and had their sights set on making the science fiction idea of air taxis a reality. (Time 0:00:29)
  • AI Art Auction Success
    • Christie’s held its first AI-only art show, exceeding sales expectations despite controversy over artists’ exploitation.
    • The top-selling piece, “Machine Hallucinations”, used ISS photos and sold for $277,000. Transcript: Nathaniel Whittemore Kitty Hawk CTO Chris Anderson is collaborating with Page again at Dynatomic’s engineering efforts. Certainly should be a fun one to watch. Next up, Christie’s Auction House has held their first AI-only art show, and sales, believe it or not, exceeded expectations. The show was extremely controversial in the art world, with 6,500 people signing an open letter urging Christie’s to cancel the event. The letter alleged, These models and the companies behind them exploit human artists using their work without permission or payment to build commercial AI products that compete With them. Christie’s Digital Art Division felt compelled to address the issue, responding, The artists represented in this sale all have strong existing multidisciplinary art practices, Some recognized in leading museum collections. The works in this auction are using artificial intelligence to enhance their bodies of work. The sale, which finished on Wednesday, offered 34 lots that sold for almost three quarters of a million dollars in aggregate. One of the pieces dated back to the 1980s and was created using a very primitive AI art program that ran on an IBM mainframe and drove physical plotters using ink and paper. Another piece was created in 1966 by digital art pioneer Charles Suri. He used a mathematical formula and computer code to distort images. The top-selling piece was called Machine Hallucinations, ISS Dreams A, by Rafiq Anadol, a pioneer in data visualization and AI art. The artwork involved feeding 1.2 million photos taken by the International Space Station into a data visualization algorithm to create an abstract video work. The piece sold for $277,000. (Time 0:01:39)
  • AI Chip Demand
    • Broadcom’s strong earnings demonstrate sustained AI demand, particularly for specialized AI chips.
    • This segment now makes up half of their business. Transcript: Nathaniel Whittemore Specialist, Nicole Sales-Giles, said, Lastly today, after the interesting art world, we’ll move over into the boring but necessary market world, where chipmaker Broadcom has smashed Earnings, reinforcing that AI demand isn’t going anywhere. The company reported $4.1 billion in AI-related revenue for last quarter, up 78% from Q4 of last year. Their numbers also showed a continued migration of demand towards AI-specialized chips. The segment now represents half of the business, up from 31% a year ago. Broadcom projected a further 7% gain in AI revenue for the current quarter, suggesting there’s no AI bust to be seen. They’ve added four additional hyperscaler clients on AI chip designs, complementing their three existing clients involved in large-centered data operation. After beating the stock down by 22% over the past month, the market loved the news. (Time 0:03:23)
  • Value Shift in AI
    • Value is shifting towards application-layer AI companies that cater to specific customer segments.
    • Data from SimilarWeb and SensorTower show which AI apps are experiencing rapid growth. Transcript: Nathaniel Whittemore Now back to the show. Do you remember when everyone was pejoratively talking about ChatGPT wrappers? Oh, it’s not a real application. It doesn’t have any defensibility. It’s just a thin layer on top of ChatGPT. Now, in certain cases, people weren’t wrong to make that critique. Although where it tended to be most applicable was the danger in starting a company based on filling in a feature that something like ChatGPT didn’t have yet, but which was clearly going To be on the roadmap. You might remember in 2023, the absolute copious number of chatbots that had PDF readers built in, which of course very quickly became just core functionality table stakes for all Of the major chatbots. Well, now there is certainly a big shift. There is a growing sense among investors in Silicon Valley that more of the value than previously thought was going to accrue to the application layer. Basically, the logic goes that companies that actually understand a particular customer segment, be they organized by some need or particular business sector, and who can design An experience that utilizes AI to solve a problem or create an opportunity for that specific buyer or specific user, are going to capture a lot of the value. The vibe shift is certainly real if you glance around from AI products being showed off on X, but luckily we don’t just have to rely on vibes. Thanks to recent reports from SimilarWeb and SensorTower, we have hard data about which AI apps are growing the fastest. These reports focus on the change in monthly active users over the past three months, so are largely focused on the hot new trends. We also have investment firm Andreessen Horowitz, who have compiled their own analysis of the top 100 Gen AI consumer apps, based both on that data but also adding additional insights From their position as one of the leading VC firms in the space. (Time 0:07:02)
  • ChatGPT Growth
    • ChatGPT’s user base doubled in six months, reaching 400 million monthly active users.
    • Growth spiked with new model releases and feature introductions, demonstrating increased engagement. Transcript: Nathaniel Whittemore Today, we’re going to dig into what is at the top of these charts, what it says about where the industry is, and what it points to for the future. First of all, if we start at the very top of the charts, chat GPT usage continues to boom. OpenAI has now doubled their monthly active users over the past six months. Looking at user growth dating back to 2022’s initial launch, you can clearly see that web traffic shows definitive spikes around the launch of new models, as well as around the initial Rollout of advanced voice mode. The long initial plateau stemmed from ChatGPT’s early novelty factor. Many consumers found it intriguing but lacked compelling daily use cases. However, as OpenAI has introduced more advanced models and capabilities into ChatGPT, usage has risen accordingly, both among existing users in a wave of new adopters. On mobile, the story is a little different, with numbers growing at a steady and ever-increasing pace. The mobile app was launched in early 2023, so we already had access to GPT-4 at that time. More than 40% of usage is now through the mobile app. Overall, ChatGPT now has 400 million monthly active users. And what’s more, the company has added 200 million of those in the past six months alone. Now, the big elephant in the room, however, for the last few quarters was DeepSeek arriving as a breakout hit. (Time 0:08:49)
  • DeepSeek’s Rapid Rise
    • DeepSeek’s chatbot quickly became a top AI product, exceeding Claude, Perplexity, and Character AI in traffic.
    • Its free, powerful model forced competitors to rethink pricing. Transcript: Nathaniel Whittemore The public chatbot generated enough traffic in its first 10 days to rank at number two on the list of top AI products, overtaking Claude, Perplexity, and Character AI. The growth numbers from SimilarWeb really demonstrate the need to wrap models in a friendly chatbot interface as a first introduction. DeepSeq was already seeing 1,000% three-month growth in web traffic at the beginning of January around the time they released their v3 foundation model. By the end of February, however, following the release of R1, growth hit 8,000%. Importantly, A16Z commented, these are not just empty downloads. Per Sensor Tower data, DeepSeq users are slightly more engaged on mobile than users of Perplexity and Claude, based on both sessions per user and minutes per user in the average week. However, they point out that engagement still significantly trails ChatGPT. One of the things that made DeepSeq so notable is, of course, that they offered a much more powerful model as a free version, which forced some changes to how other companies were thinking About this as well. (Time 0:09:58)
  • Sector-Specific AI Growth
    • AI tool web traffic growth slowed, but sectors like data analytics, HR, and coding assistance surged.
    • This highlights a shift towards specialized AI solutions. Transcript: Nathaniel Whittemore One of the interesting lenses on the similar web data is a breakdown by segment. Overall, web traffic for AI tools grew at 20% for the first three months, ending in February. This growth has slowed down a little from the 38% registered in November. The sectoral analysis, however, is very lumpy. Some segments like customer support, music generation, and writing and content are in outright decline. Others are really hitting their stride. Data analytics saw 42% three-month growth by the end of February, the fastest pace in September. Human resources grew at 31%, while legal grew at 12%. I think that this says a lot about the shift towards purpose-built verticalized AI solutions, which it seemed like had been a theme, but these numbers really show that to be the case. The fastest growing segment, unsurprisingly, was coding assistance. Growth for this segment accelerated profoundly at the beginning of this year and surged to a high of 83% in late January. (Time 0:10:54)
  • Coding Assistance Boom
    • Coding assistance tools experienced explosive growth, splitting into agentic IDEs (like Cursor) and text-to-code platforms (Bolt, Lovable).
    • These tools empower non-coders to build functional apps quickly. Transcript: Nathaniel Whittemore It was still surging ahead at 72% by the end of February. A16Z broke this segment down into two categories, both having a huge moment. The first was agentic IDEs designed for advanced programmers. Cursor is, of course, the flagship app for this use case, which if you want an idea of how fast things are changing, this is actually the first time Cursor has featured on A16Z’s top 50 Web apps. The other big breakout is Vibe Coding. In other words, a prompt-based coding where you talk to a tool that does the coding for you rather than manually typing the code yourself. Thanks to these text-to programs, Replit has that functionality, but some like Bolt and Lovable are both purpose-built for that. With these programs, we’re at the stage where a non-coder can spin up a functional app in minutes. Or as we use it in Super, entire teams of non-technical people can prototype ideas and share clickable, interactable prototypes, rather than just trying to type up different feature Ideas or share design mockups. Bolt reached $20 million in annualized revenue and $2 million registered users in its first two months. Lovable has reported $17 million in annualized revenue across their first three months. Comparing these two sectors, the top two agentic IDEs are outpacing the top two text-to tools, but it’s pretty close. Each segment is showing stratospheric growth since the middle of last year. Fascinatingly, but perhaps not unexpectedly when you actually think about it, the overlap between users of these two types of tools is relatively small. According to SimilarWeb, just 23% of Bolt’s users also visited Cursor in January. Could represent that these really are functionally different audiences. Part of this shift in ViveCoding is being driven by all these people sharing their interesting examples. Peter Levels, for example, showed off a flight simulator that went viral. The infamous solopreneur cobbled together an online multiplayer game with minimal coding knowledge and a Cursor subscription in the space of just a few evenings. He’s been adding features and fixing bugs, using Gen.ai over the past two weeks, self-admitting that he has no idea what he’s doing. By selling in-game advertising, he’s managed to build up $67,000 in MRR in 13 days. Now, it might be a little bit of a stretch to call this actual monthly recurring revenue given this is a viral hit. It also doesn’t hurt that Levels has a significant distribution funnel and was retweeted by Elon Musk, but it still demonstrates the point and the aspiration of these new tools. (Time 0:11:45)
  • Revenue in AI Apps
    • AI content production apps lead in revenue generation, while some niche categories, like nutrition and language learning, also show promise.
    • Niche apps can successfully attract paying users. Transcript: Nathaniel Whittemore Another interesting part of the A16Z study was their mapping of the difference between apps that are getting a lot of users and the ones that were generating a lot of revenue. Setting aside ChatGPT, which is dominating on both levels, AI content production apps were the clear winner. In video editing apps, they noted that the top three by monthly active users were completely separate from the top three by revenue. They also noted that some categories were ranking for revenue but had very few users. Plan identification, nutrition, language learning, music generation, and dictation. Each had either one or two apps that had cornered the market, suggesting that niche AI apps can develop a loyal user base who are willing to pay for tailored services. (Time 0:13:56)