What Happens When AI Obliterates Your Business Model?

To Dan the AI Daily Brief, what happens when AI undercuts your business model even when your product is doing great? Before that on the headlines, is our AI personal assistant going to live in our email inbox? The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. Alright, friends, quick announcements before we dive in. Like you first, as always, to our wonderful sponsors today, that's KPMG Zencoder and Super Intelligent. If you would like a blissfully ad-free version of the show, you can get that at patreon.com/aidlybrief or you can subscribe on Apple Podcasts to learn about sponsoring the show. Check out aidlybrief.ai. It's bullet three. There's also an email there. And while you're there, there's so many other things that you can check out as well. I announced yesterday that we've got this new strategy compass platform coming and I am signing up people for the beta. You could also find the link to our AI New Year's resolution community, we're on week two of ten. Or you can find out more about AIDB Intel. I've got some big announcements on that front coming up, so definitely sign up there at aidbintel.com. Basically, if you want to learn anything about the podcast and its broader ecosystem, aidlybrief.ai. And with that out of the way, let's talk AI. Welcome back to the AI Daily Brief headlines edition, all the daily AI news you need in around five minutes. One of the great white whales of the AI space is the quest to build the AI personal assistant. So many people are thinking about how they become the default personal assistant and what that means in terms of their ability to retain users. And with its latest move, Google is pointing out that maybe the place for that to live is just about the most obvious place you can imagine. This week, Google announced a big AI overhaul for, you guessed it, Gmail. First up, Google is bringing AI search to the inbox. Users can now search using natural language rather than keywords. Google gave the example of being able to ask, "Who was the plumber that quoted for the bathroom innovation last year?" The response will then be an AI overuse style summary with links to relevant information, saving users from hunting around for key pieces of information. There's also two new tabs labeled suggested to-do's and topics to catch up on. These features both scour your emails to populate useful lists. Suggested to-do's will detect things you might need to take action on, like a bill coming do or an appointment to confirm, while topics to catch up on will give you a passive list to keep you up to date on things like e-commerce order tracking or financial statements. In addition, Google is moving more Gemini features from the paid tier to their free product. All users can now use the help me write feature to polish emails or generate first drafts. There's also an AI-generated suggested replies feature that can offer up one click responses based on your writing style. I've also noticed that outside of just trying to imitate your style, it can get the context of emails where you repeat yourself a lot and simply suggest them up front. I have a very standard response I send back, for example, to sponsorship inquiries, and at this point, Gmail knows to just suggest that whenever there's a sponsorship inquiry. Speaking to the AI-generated lists, Blake Barnes, the VP of Product for Gmail said, "This is us delivering on Gmail proactively having your back, showing you what you need to do and when you need to do it. Don't worry, the traditional inbox will remain available. This is simply a new view you can toggle in and out of as you please to cut through the noise of your incoming mail." Now honestly, I've been a little bit surprised by the response, which has been kinda positive. Maybe I'm just jaded and expect everyone to hate everything instantly, but what you're seeing is responses like this one from Koho Akada, who writes, "If this helps me stop missing bills and important emails I'm all in, AI that actually saves time over everything else." Trump fan wrote, "Gmail's new AI inbox feature automatically creates a to-do list for you based on important email and calendar events. If this actually works well, inject it directly into my veins." So I don't know, man, I guess we gotta go try this out now. Next up, a bit of chip and geopolitical news, Nvidia is ramping up production of AI chips for China, but Beijing is throwing on the brakes. During a Tuesday Q&A at CES, Jensen Huang said that Chinese demand for H-200 is quite high after the Trump administration announced the advance chips would be available for export last month. He said, "We fired up our supply chain and H-200s are flowing through the line." Huang said that Nvidia is in the "his words final innings of sorting out the paperwork in Washington, but on the Chinese side," he said, "we learn about everything through purchase orders. We're not expecting any press releases or large declarations." On Wednesday, however, the information reported that those purchase orders are getting pulled. Beijing has reportedly told tech firms to halt orders while the terms required to access chips are decided. The Chinese government is of course stuck between allowing the H-200s into the country, giving their tech company's access to more powerful chips, or on the other hand, keeping American chips out of the country in order to bolster domestic chipmakers like Huawei. The issue is that the chips that are available domestically are still fairly far behind. The report states that Beijing is expected to mandate domestic chip orders in exchange for approving Nvidia imports. Liu Peng, U.S. spokesperson for the Chinese Embassy in the U.S., said, "China is committed to basing its national development on its own strengths and is also willing to maintain dialogue and cooperation with all parties to safeguard the stability of global industrial and supply chains." Later in the day, Reuters added reporting that Nvidia is asking for cash in the bank to secure Chinese orders. Sources said that they're demanding full payment in advance to hedge against uncertainty and Beijing's approval process. The unusually strict terms also include no option to cancel, no refunds, and no ability to alter configurations once the orders are placed. These sales are a huge financial commitment for Nvidia, and Beijing's approvals could make or break their year. They already have 700,000 H-200 in inventory, but reports suggest more than two million Chinese orders have already been placed. What's more, H-200 demand from the rest of the world is now close to zero now that blackwell chips are available, so additional production is a big risk. Last year, Nvidia guided that Chinese demand could add between $2 billion and $5 billion to their quarterly revenue on the order of a 5% bump. However, at $20,000 a pop, $2 million orders is closer to $50 billion in revenue around 40% of their full-year revenue for 2025. Staying on Beijing for a moment, the Chinese government is also disrupting Meta's plans as they consider intervening in the Manus acquisition. Over the winter break, Meta announced the $2 billion acquisition of Manus, a big part of the narrative was that Manus had figured out a playbook for Chinese startups to court lucrative deals from US tech companies. While Manus was founded in China, they relocated to Singapore last year and an apparent plan to avoid regulatory issues from both Beijing and Washington. It seems that that plan has not quite worked the way that they had hoped. The Financial Times reports that the Chinese Commerce Ministry is assessing whether Manus's relocation and subsequent sale require a technology export license under Chinese law. Sources said the review is still in the early stages and might not lead to formal investigation. However, the license requirement could give Beijing a way to influence the deal, up to and including forcing it to be unwound. The South China Morning Post reported on the story in the context of a broader trend, called Chu Hai, are going to see where Chinese tech firms and startups establish overseas subsidiaries. They write, "Beijing has been exploring ways to assert a say in cross-border transactions that involve Chinese technology data, talent, or markets." Sources said then that allowing the Manus deal to go through is viewed as a dangerous precedent that could encourage more Chinese startups to relocate offshore. Shifting over to AI and Commerce, Salesforce reports a huge uptick in AI shopping in their annual holiday survey. This holiday season saw a large increase in sales compared to 2024, with 12% growth globally and 9% in the US. Other data showed that this was generally about inflation, with the increase largely seen in the value of goods purchased rather than an increase in volume. Sales force viewed this shift in consumer behavior as being about shoppers being more discerning about how they spend their money. And the thought is that AI is playing a major role in helping consumers make decisions. The holiday saw 20% of all retail sales powered by AI in agents, either through recommendations or agentic shopping. The share of traffic from chat Gbt and perplexity doubled this year and conversion rates were way up. AI searches were 9 times more likely to convert into sales than social media referrals. Kayla Schwartz, the Director of Consumer Insights at Salesforce, said, "One of the biggest stories of the 2025 holiday shopping season is consumers leaning into these tools, both on-site branded agents as well as off-site third party agents, and seeing how AI was really aiding and abetting the other aspects of the shop or journey, particularly within customer service." Now on this theme of shopping, while many AI companies are building shopping features, Amazon seems to be getting out to an early lead. Earlier this week, there was a controversy around Amazon displaying products available from small retailers who don't stock their products through Amazon. Some retailers objected, stating that Amazon didn't have approval for the listings. Others mentioned errors with the listings, including out-of-date product information or out-of-stock items. It turns out the listings were part of a new feature called Buy for Me, which uses AI agents to buy the products from third party websites. Amazon appears to have obfuscated the entire process, presenting customers with a single button that they wouldn't even know as an AI feature. What's interesting is that this is a form of agentic shopping that open AI and others had envisioned last spring as the features were coming to market. The original plan was to have a web agent able to gather products from anywhere on the internet and serve them up in a chatbot session. That approach ran into snags with many online shopping websites blocking the crawlers that enabled it. That led instead to open AI partnering with large retailers like Shopify and Etsy for data access. Chief among the sites blocking web crawlers was Amazon, obviously not keen on their customers buying through chat GPT instead of their own frontend. It seems that Amazon is using their leading online marketplace as a mode while also using agentic shopping to tap into demand for smaller retailers. Ironically, the Buy for Me feature allows Amazon to tap into Shopify stores without needing to partner directly. For the moment, Amazon seems to have the most complete and functional agentic shopping experience. And it's completely housed within the standard Amazon website that people are already comfortable using. Right the information by taking a more collaborative and slow rolling approach, open AI and Shopify could risk seating ground to Amazon, especially since it's still early days of people getting used to the idea of AI doing their shopping. One more on this topic, Microsoft has added agentic shopping to their co-pilot experience as well. A new feature called co-pilot checkout was unveiled at the NRG retail conference on Thursday, allowing users to shop without leaving the app. Like other AI shopping features, consumers can use the chatbot to perform product research and compare prices then check out without ever leaving the app. The feature is now live in the U.S., and includes integrations for Shopify, PayPal, Stripe, and Etsy. Microsoft is also reinforcing the narrative that chatbot conversations seem to have much greater intent than other forms of online interaction. Microsoft said that shopping sessions that include co-pilot led to 53% more purchases within the first 30 minutes than those without AI. In addition, they said that shopping sessions using co-pilot are 194% more likely to result in a purchase. I actually talked about this AI commerce theme on someone else's podcast that will be coming out in a couple of weeks. I think it's quietly one of the big trends that's happening right before our eyes, even if we're not talking about it all that much. For now, though, that is where we will wrap the headlines. Next up, the main episode. Alright, let's talk about the signal versus the noise in Enterprise AI. The challenge right now isn't just about what's possible, it's about what's practical. That's the entire focus of the UCan with AI podcast I host for KPMG. Season one, cut through the hype to focus on deployment and responsible scaling. Season two goes a level deeper. We're bringing together panels of AI builders, clients, and KPMG leaders to debate the strategic questions that will define what's next for AI in the Enterprise. Six episodes packed with frameworks you can actually use. Find UCan with AI wherever you get your podcasts. Subscribe now so you don't miss the new season. If you're using AI to code, ask yourself. Are you building software or are you just playing prompt roulette? We know that unstructured prompting works at first, but eventually it leads to AI slop and technical debt. Enter ZenFlow. ZenFlow takes you from vibe coding to AI first engineering. It's the first AI orchestration layer that brings discipline to the chaos. It transforms freeform prompting into spectrum and workflows and multi-agent verification, where agents actually cross-check each other to prevent drift. You can even command a fleet of parallel agents to implement features and fix bugs simultaneously. We've seen teams accelerate delivery to x to 10x. Stop gambling with prompts. Start orchestrating your AI. Turn raw speed into reliable production grade output at zenflow.free. Today's episode is brought to you by Super Intelligent. Super Intelligent is a platform that very simply put is all about helping your company figure out how to use AI better. We deploy voice agents to interview people across your company, combine that with proprietary intelligence about what's working for other companies, and give you a set of recommendations around use cases, change management initiatives, that add up to an AI roadmap that can help you get value out of AI for your company. But now we want to empower the folks inside your team who are responsible for that transformation with an even more direct platform. Our forthcoming AI strategy compass tool is ready to start to be tested. This is a power tool for anyone who is responsible for AI adoption or AI transformation inside their companies. It's going to allow you to do a lot of the things that we do at Super Intelligent, but in a much more automated, self-managed way, and with a totally different cost structure. If you are interested in checking it out, go to aiDailyBrief.ai/compass, fill out the form and we will be in touch soon. Welcome back to the AIDailyBrief. For those of you who are not coders and not technical, there is actually a cheat code that will allow you to live at least a little bit in the future, at least when it comes to understanding AI, how tools are going to be used, what new processes are taking hold, and what disruption to knowledge work looks like. By watching what actual software engineers and developers who are deep in the AI space are talking about, thinking about, and struggling with, is as close to living in the future as you can get. Now, I talked about this a little bit in end of year episodes when it came to some of my vibe coding predictions. In short, I think a lot of other sectors of knowledge work outside of coding will have their work vibified in 2026 in the way the developers did in 2025. But it's not just positive change that we get to preview. In many ways, even as software engineering is the area where AI is enabling the greatest leverage that didn't exist before, it's also the area that is most struggling with disruption and change to what the very essence of the profession means. That is playing out on both an individual level, whereas we saw in the cloud code episode from earlier this week, some developers are struggling with the existential question of what it means to be a developer in this new paradigm, but it's also playing out on a business and organizational level. Here this week, the story of Tailwind went viral. Tailwind maintains an open source CSS framework that allows web developers to speed up front and development. It is extremely popular among developers and among AI coding tools. Tailwind's model is free and open source, with a paid plus tier that drives revenue. Where the story started to pick up was actually from a comment on GitHub. A user was asking for a text-only version of the documentation to make it more easily ingested by AI. CEO Adam Wathen responded, "I totally see the value in the feature, and I would like to find a way to add it, but the reality is that 75% of the people on our engineering team lost their jobs here yesterday because of the brutal impact AI has had on our business. And every second I spend trying to do fun free things for the community like this, is a second I'm not spending trying to turn the business around and make sure the people who are still here are getting their paychecks every month." He went on to note that traffic to their documentation is down 40% compared to 2023, despite Tailwind being more popular than ever. This is the problem because as Adam put it, the docs are the only way people find out about their commercial products. And without customers paying for that plus tier, they can't maintain the framework. Netnet, while Tailwind was growing faster than it ever had and was bigger than it had ever been, their revenue was down close to 80%. So the width of this subverts the narrative of AI disruption, is that this is an AI undermining the need for something, the product is more popular than ever. The problem is just that the AI doesn't need the documentation that leads customers to the paid product. And while open source software being difficult to monetize and build a startup around is nothing new, it was this unique Hatch 22 of AI adoption, where AI was driving more usage but driving down revenue that had people sit up and take notice. Santiago tweeted, "Tailwind laid off 75% of their team. At a time when Tailwind is more popular than ever, their revenue is down close to 8%. LLMs did this. If we don't figure this out, we'll end up with a massive graveyard of abandon wear." Yash Barge was summed up, their CSS framework became extremely popular with AI coding agents, 75 million downloads a month, that meant nobody would visit their docs where they promoted paid offerings, resulting in a 40% drop in traffic and 80% loss in revenue. Nick Papa George writes, "Tailwind is a canary. AI can work with a greater open source version of anything now. So there's no need for the premium dev-friendly DX Forward version. Auth databases on and on will change things quickly in my opinion." Valentining day shift writes, "Making a business on top of open source software was already near impossible, and now OSS maintainers who want to earn money are extra screwed. AI will scrape your project site, users will never visit it for documentation, and will never know about your commercial product. And we're not talking about some random small library. This is the world's most used CSS framework that's growing day by day. At this point, the only thing you'll be able to do in open sources upselling LLM tokens." Now this particular story actually ends at a more positive note. On Wednesday, CEO Adam recorded a podcast on his morning walk discussing the situation whose company was in. People responded to it for its openness and transparency, and the post ended up with almost two million views. By that evening, the donations had started to come flooding in. Logan Kill Patrick announced that Google AI Studio had become a sponsor of the project, Antoine Asica from Lovable writes, "Every app built on Lovable uses Tailwind and we owe them a lot. We're now sponsoring them and I genuinely encourage fellow founders to do the same." And the fellow founders piled on, "Superbase added themselves to the list, saying we're excited to become a Tailwind partner and will continue to support their tremendous contributions to the industry." Garema the CEO of Versels said, "Versel will be officially sponsoring Tailwind CSS. That's a given." Tailwind is foundational web infrastructure at this point. I've also reached out to Adam to explore how we can make this a longer term commitment. Gumroad cursor, the list goes on and on and on. The challenge though is that this isn't necessarily an isolated incident. Another interesting story from this week is that of Stack Overflow. For the non-programmers in the audience, Stack Overflow is a forum that's been around since 2008 to answer programming questions. If you think about it as Reddit for programming, you're roughly on the right track. Throughout the 2010, Stack Overflow was pretty close to critical infrastructure for the tech industry. It could help beginners figure out straightforward issues, but it also had a deep user base of incredibly knowledgeable coders. If you add an expert problem with a barely used language, it was a pretty good bet that the small handful of people in the world that could answer your question could be found on Stack Overflow. Even before ChatGPT launched, Stack Overflow seemed to have reached its peak. However, there is absolutely no denying, that when ChatGPT launched, it completely aviated the need for the site. In short, there was no point in trawling through forum posts when all of that was to be found somewhere in the AI training data that could be surfaced with the natural language query. All the way back in July of 2023, Elon Musk called Stack Overflow's plight Death by LLM. Now the reason that the slow motion demise of Stack Overflow is back in the news this week is that it finally reached the end. Last month, the forum registered 6,866 queries, roughly the same as their first month way back in 2008. That's down from a peak of 300,000 queries a month at its top in 2020, and it's consistently more than 150 to 200,000 that it had for almost a decade before that. And so the question for all of this is what it adds up to. In the case of Stack Overflow, there's an interesting broader implication of what happens when there's no longer a need for massive open repositories of human knowledge. If there's no longer a culture of building giant forms full of human expertise on the open web, that creates a challenge even for AI. As effect fully puts it, Stack Overflow's downfall will send shockwaves far into the future because there will no longer be a source of high quality, well-structured data to train AI on. But what about when it comes to tailwind and the dramatic shifts to business model that AI represents? And to be fair, although tailwind CSS is a beloved product, some people pointed out that it was never a super solid business. Daniel Jeffries wrote, "I'm having a lot of trouble buying this AI killed my business narrative for tailwind, although I feel for the excellent team who does incredible work. It took me a bit of hunting to even find that tailwind has products by digging around their site, and a good marketing person will tell you to make those offerings front and center. Even in the age of AI-centric coding, my team buys a lot of SaaS, that awareness problem can be fixed with a good marketing person who makes the offerings jump out. And speaking of those offerings, it also looks like the offerings are one-time purchase so there's no compounding revenue, and it's also supported by donations. Having lived off donations with a nonprofit, I felt first hand what it's like to go from two million in donations to practically nothing overnight because the market shifted. In short, AI didn't kill tailwind, AI exposed a fragile business model. Docs traffic down because LLM's answered questions directly. UI kits down because LLM's generate UI instantly. One time purchases down because there's no recurring value. Popularity up, but value capture near zero. It's not an AI governance failure, that's no mode and no recurring revenue. Open source has faced this exact problem many times. MongoDB fixed it with Atlas, Elastic fixed it with hosting services, GitLab fixed it with enterprise tiers. Tailwind just hasn't crossed that bridge yet. I hope this fine team finds their way, but the AI killed by business is a bit off the mark in more ways than one, and our willingness to just accept this story at face value these days is depressing. Akash Gupta explored this as well. He wrote, "What's the fix? The path forward probably requires building something AI can't replace. Services, consulting, enterprise contracts, deep integrations that require human judgment, or build features that actually get better with scale instead of just hoping people stumble on your pricing page." But as Akash points out, the whole situation is a preview of what's coming for information businesses generally. If your value is answering questions that AI can now answer, your mode just vanished. And that I think is the interesting consideration. This might be happening in the context of development now, but it will have similar seismic effects in other areas in the months and years to come. For my part, I think it'll be interesting to see what comes next. Tailwind represents this interesting category that's almost like a public good at this point. Certainly, the way that companies rose up to try to support it suggests for that interpretation. One possibility is suggested by biology Shrinivasan, one of the big AI companies should consider requiring tailwind or do a strategic investment and maybe rehire all the devs. They've given the ecosystem so much. In other words, basically a big patron just secures the public good for the market for perpetuity. Another approach was proposed by Nat Eliasin who wrote, "With the tailwind news, it would be cool if Cloud Code let me pay 1% more to automatically contribute to the open source projects I use based on my token spend." Or they could just do it. So will we see a different conception of public goods and digital infrastructure arise? I'm not sure, but like I said at the beginning, if you want to get a preview of what everyone else is going to be dealing with six months from now, there's basically not much better you can do than watching what developers are talking about right now. For now, that's going to do it for today's AI Daily Brief. Appreciate your listening or watching, as always, and until next time, peace.