The Genius Who Scaled PlugAI to $200K/Month, Umax to $6M ARR, and Cal AI to $50M ARR — Then Sold It Inside Blake Anderson's playbook — the 30-minute idea test, the exact influencer math, and how a failed reality show turned into a studio run by teenagers. Blake Anderson didn’t just build three hit apps. He locked himself in an apartment, failed on camera, and turned that failure into a studio that puts teenagers in the CEO chair. This newsletter breaks down real-world cases of people making serious money with apps in the AI era. Today’s subject: Blake Anderson. He’s back — the sharpest young app founder in the game, and honestly, many successful app builders I’ve researched point to him as an inspiration. This is actually the second time I’m covering him. I first wrote about him back in October 2024, when I was floored to learn that a guy with zero engineering or product background had built a portfolio of apps generating hundreds of thousands of dollars a month, just by talking to ChatGPT. That was impressive enough at the time. But what’s happened since then is on another level entirely: - His co-founded app, Cal AI, crossed $50 million in annual recurring revenue and sold to a major acquirer - He made Forbes 30 Under 30 - He now runs an app studio for young developers called “10x” And in the middle of all that, he also locked himself in an apartment for 90 days as part of a reality show chasing $1M in annual revenue — and failed, publicly and completely. Blake puts both his wins and his losses on full display as he keeps evolving. This piece is a full look back at his turbulent two years. First, a Quick Recap of Blake’s Earlier Chapters Let’s start with the basics. His career began with three apps. The first was PlugAI (originally called RizzGPT) — upload a screenshot of your dating-app conversation, and the AI generates a reply designed to land. It was born from his roommate asking, “What do I say back to this girl?” With zero software engineering experience, Blake built it in two weeks by asking ChatGPT how, at every step. It went viral on TikTok, hit 200,000 downloads in six days, and crossed $200,000 in monthly revenue within a few months. The second was Umax — riding the “looksmaxxing” trend among Gen Z men, where an AI scores your selfie and hands you a personalized self-improvement plan. It hit $100,000 in its very first month. From there, powered by aggressive influencer marketing, it went viral and scaled to $6 million in ARR in three and a half months. And the third was Cal AI — a calorie tracker that works by just photographing your meal. Blake co-founded this one with two high schoolers, then 17 years old: Zach Yadegari and Henry Langmack. That’s where my previous article left off. Everything below is what happened after. Cal AI Turns Into a Monster At the time of my last article, Cal AI was described as “recently launched, apparently just crossed $100K a month.” The growth from there has been genuinely absurd. By March 2025, Cal AI had passed a million cumulative downloads and got covered by TechCrunch. By September 2025, CNBC ran a feature on it. And eighteen months after launch, it crossed $50 million in ARR and 15 million cumulative downloads. Blake shared his July 2025 monthly revenue on X — that month alone, his app portfolio pulled in a combined $3.9 million. The breakdown: Cal AI brought in $3.31 million, PlugAI added $360,000, and Umax — well past its peak by this point — still contributed $270,000. Two years ago I introduced him as “the guy making $600K a month.” He’s now running north of six times that. Cal AI’s CEO, Zach Yadegari is a 17-year-old who scaled an app to $8.5M in monthly revenue in under two years and then sold it. What makes this team interesting is how clean the division of labor was. Zach ran the whole thing as CEO. Henry handled the engineering. And Blake injected the “how to make TikTok go viral” playbook he’d built through RizzGPT and Umax directly into the team. Teenage momentum, plus the playbook of a guy who’d already hit twice. That combination is what produced a monster app. Cal AI Sells to MyFitnessPal Then, in early 2026, the biggest news of all landed: Cal AI was acquired by MyFitnessPal, the giant of the nutrition-tracking category. According to TechCrunch’s reporting, the deal closed in December 2025 after nearly a year of negotiations. The purchase price was never disclosed — but for an app doing $50M in ARR with 15 million downloads, it’s safe to assume it was significant. A guy with zero engineering background, who built apps by asking ChatGPT how, ended up with a full exit on a co-founded company in a little over two years. Looking at a story like this, it’s easy to conclude Blake is simply a statistical outlier. And sure — he’s clearly talented. But the way he got started in the first place was as ordinary as it gets: throwing his roommate’s dating-app problem at ChatGPT. Anyone could have taken that first step. The “Lockdown Reality Show” — and a Public Failure If the story stopped here, it would read as a clean success story. But Blake’s last two years weren’t a straight line. Back in October 2024, he made a declaration that surprised people watching him: “I’ve run this exact playbook three times in one year, idea to execution. I’m done doing it this way. Ask me anything — I’ll share all of it.” He was walking away from the exact formula — RizzGPT, Umax, Cal AI, three hits in a row — that had defined his career. Then, in February 2025, he threw himself into something wild: a show called PMF or Die. The show was produced by the team behind TBPN (originally launched as “Technology Brothers”), the popular tech podcast that was later acquired by OpenAI and made headlines in its own right. The rules were simple and brutal: - Two founders get locked in a roughly 700-square-foot New York apartment - Food, drinks, and everything they need are provided - They start with $25,000 in cash and an internet connection - They have 90 days to build a business from scratch that hits $1 million in annual revenue - They can’t leave the apartment until they hit the number - The entire thing streams live, 24 hours a day, 365 days a year It’s an unhinged premise — and it worked exactly as intended, blowing up online. Blake personally recruited a 21-year-old developer, Patrick Callaway, and the two of them moved into the apartment together. In his launch video for the show, Blake laid out the real point of the exercise: “I’ve talked publicly about making millions of dollars off these apps. A lot of people call it luck. This is about proving it’s repeatable.” The stakes got real enough that Polymarket, the prediction market, opened betting on whether they’d succeed or fail — money from around the world riding on the outcome. Inside the apartment, the two of them built a learning app — also called “10x” — that generates AI lessons on any topic a user wants to study, and submitted it to the App Store on day 31 of the challenge. So — how’d it go? The result: failure. A little over 30 days into the challenge, Blake quit. Not because the business wasn’t progressing, but because being confined to that apartment, under constant surveillance, simply became unbearable. It’s a disappointing outcome, but honestly, an understandable one. Hitting a $1M revenue target is one kind of challenge. Being locked in an apartment for 90 days under 24/7 livestream surveillance is a completely different kind of pressure — one that has nothing to do with whether the business idea itself is good. But what he did after the failure is what defines the Blake Anderson of today. He kept the name “10x” and rebuilt it into something entirely different. Instead of building apps solo, he now recruits talented young developers, injects his own playbook and capital into them, and lets each one launch and run their own app. Blake moved from being the player to being the one who provides the environment, the capital, and the know-how. A guy who, as an individual operator, had already built apps generating tens of millions of dollars, now switched to “mass-producing the type of player who can do that.” In a real sense, Cal AI’s success can be seen as the first proof point for this exact studio model. So from here, let’s get into the specific playbook he now teaches — publicly, in full. The Fully Public “Viral App Playbook,” Broken Down As mentioned, right after declaring he was done running his old formula, Blake started publishing the entire playbook behind it. Here’s what’s actually in it. Big Problem × Simple Solution Start with idea selection. His core principle: attack a big problem with a simple solution. The categories he recommends going after: - Health - Dating/relationships - Career - Education - Self-discipline - Addiction - Money The categories he says to avoid: - Social - Entertainment - Niche hobbies - Crypto He’s especially emphatic about one rule in particular: never build a social app. The common mistakes that show up right before a developer’s breakout success. Blake himself failed at a marketplace startup back in college. It’s a consistent pattern: CtoC platforms are territory that under-capitalized solo builders shouldn’t be trying to win. Idea Validation: The 30-Minute Rule Next is how he actually validates an idea — and this part is genuinely useful. The moment he has a new app idea, he creates a brand-new social media account. Then he follows nothing but content related to that theme and spends 30 minutes a day, every day, immersed in that feed. Do this consistently, and the target user’s entire mindset installs itself in your head. What language do they respond to? What are they actually struggling with? What video formats are going viral in that niche? Both the app’s feature set and its marketing angle start to emerge naturally from that immersion. Before writing a line of code, he’ll also test the idea by posting content on the topic and watching the reaction. That gives him a validated signal before the product even exists. The pattern among winning builders overseas isn’t “build it, then try to sell it.” It’s “know it’ll sell, then build it.” Design Rule: “Usable by a 7-Year-Old and a 70-Year-Old” His design philosophy is just as blunt: - A user needs to understand what the app does within two seconds - Cognitive load has to be low enough that a 7-year-old and a 70-year-old can both use it without friction - Before designing anything, paste screenshots of every competitor app into Figma first It checks out — every app he’s touched is dead simple, with value that’s instantly obvious. Build in Shareability The other critical feature ingredient is shareability. The screen itself needs to be designed so it works as native content on TikTok or Instagram. Umax’s “your face gets scored on screen” moment is a perfect example — that single screen is, by itself, a ready-made short-video hook. You want features that naturally spawn user-generated content, and that has to be baked into the core product from day one, not bolted on later. When the screen itself works as content, pitching influencers gets dramatically easier, and organic UGC starts happening on its own. Marketing, in other words, starts at the design stage — before the app is even built. Use Influencer Marketing Aggressively Now for the marketing section, which is genuinely the most tactical part. Blake organizes marketing into three channels: - Organic viral content (the only option if you have no budget) - Sponsored content from influencers - Paid ads (only once your LTV supports it) The influencer marketing process is the most specific — and the most actionable. At zero traction, he DMs a carefully selected list of 60 micro-influencers, aiming to land partnerships with about 20 of them (a 33% hit rate). The key moves at this stage: - Create and show them a mocked-up example of the promotion yourself - Explain the app’s core feature in one sentence - Offer $50–$100 per post, paid immediately, to remove all friction - Don’t dictate the format — let each influencer run it in their own style Once an early winning pattern emerges, he automates outreach with AI agents and shifts the compensation model to a hybrid of flat fee plus performance-based payout. On PlugAI specifically, he went a step further and built a dedicated community for affiliate influencers, where they could share which viral ad formats were working with each other. He effectively turned the marketing channel itself into a self-sustaining community. This isn’t unique to Blake, either — plenty of the successful apps I’ve covered in this newsletter run the same play. PingoAI, for instance, runs a program called “Pingo Creators” for exactly this purpose. Reinvest Aggressively — and Decisively Blake is also known for something else: he doesn’t pocket the money he makes. He puts it right back into aggressive bets. The clearest example is the “MrBeast play” executed through Cal AI. The team put $500,000 into a sponsorship slot with MrBeast, the world’s biggest YouTuber. On the surface, that’s an insane number for an app at their stage. But their math looked like this: they were confident they’d recoup at least $300,000, with real upside into the millions if it hit. And even in a worst case where they lost the entire bet, they calculated they could earn it back within one to two months through the rest of the business. Beyond the direct ROI, the deal did something harder to measure: it bought them the credibility of being “the app MrBeast endorsed,” which opened doors in every subsequent influencer negotiation they had. Umax ran the same playbook under pressure — when competitors started catching up, the team poured over $200,000 straight into influencer marketing in one push and won the category outright. The principle: once you can see the winning path clearly, bet the maximum you can afford to lose. It’s worth noting MrBeast himself is famous for doing exactly this — plowing his own earnings straight back into the next, bigger production. The people who break through tend to share this trait: they don’t stop at making money. They put it directly into the next swing. A New Model: Handing the Company to Teenagers Now, back to the “10x” studio mentioned earlier. After the PMF or Die failure, Blake rebuilt “10x” into something completely different: he recruits young developers, gives them capital and his playbook, and lets each one launch and run their own app. One standout from that studio: a college student named Benji, who built an app that reached $30,000 in monthly revenue within a few months. His story is worth its own piece. He’s a 20-year-old student in New York who’s built more than 45 apps over the past year. His app inside the 10x studio, “Snag” (an app for finding free items available nearby), hit $30,000 in MRR in under four months from launch. The Parent Company — and a 15-Year-Old-Run Spinout Today, “10x” has evolved into an interesting two-layer structure. The parent company is 10X (10x.so), based out of an office in SoHo, New York, with Blake as CEO. The COO of that parent company is none other than Benji (Benjamin Chen) — the 20-year-old student who scaled “Snag” to $30K in monthly revenue inside the studio is now, effectively, the company’s number two. That’s how the 10x development model actually functions: perform inside the studio, and you move up into the leadership team. And the concrete product manifestation of “10X’s builder-facing AI software” is the spinout project, 10x (10x.app) — an AI app builder pitched as “Shopify for mobile apps.” Describe the app you want to build in plain language, and the AI generates native SwiftUI code and handles the App Store submission for you. It even analyzes competitor apps’ ad performance, paywalls, and monthly revenue, and sets up a social media presence for your app the moment it’s built. It’s essentially Blake’s entire “idea to monetization” playbook, rebuilt as software. Pricing starts at $20/month. And here’s the part that’s genuinely hard to believe: look at 10x.app’s leadership team on their official site. - CEO: Evan Yadegari, age 15 - CTO: Timmy McKeegan, age 16 - CMO: Albie Charven, age 14 - Partner: Blake Anderson The CEO is 15. And he’s the younger brother of Zach Yadegari — Cal AI’s own CEO. At 14 years old, he scaled a screen-time management app called “Locked” to $14,000 a month. So here’s the full picture: Blake runs the parent company, 10X, as CEO, providing capital and the playbook. Underneath it, a team of teenagers runs their own product. 10x.app, under Evan’s leadership, reportedly hit $30,000 in ARR within its first 30 days of launch. The older brother, Zach, teamed up with Blake to build Cal AI and sold it at a $50M ARR valuation. The younger brother, Evan, is now under Blake’s umbrella too — but instead of building apps, he’s building the tool that builds apps. The Yadegari brothers and Blake Anderson, working together in two different generations of the same playbook — it’s a genuinely absurd combination. And when you think about it, this structure makes total sense. The heart of Blake’s playbook is “grow youth-facing apps through TikTok virality.” Who understands Gen Z’s instincts better than Gen Z itself? Put teenagers in charge, and back them with capital and know-how from the parent company. That’s what “10x” actually is today. And on reflection, this positioning is arguably the strongest one Blake could occupy. Hitting on your own apps repeatedly is high-variance — a matter of luck as much as skill. But “someone who knows the formula, giving talented young people the environment to run it themselves” is a genuinely repeatable business. Zach and Henry, the team he built Cal AI with, were themselves “talented teenagers” first. Blake essentially turned his own success pattern into a company. Closing Thoughts So that’s Blake Anderson’s last two years. The list of app founders who’ve been directly influenced by him is long — not just Zach and Evan, but a huge share of the founders I’ve profiled in this newsletter cite him as an inspiration for how they grew their own apps. And he keeps challenging himself, keeps stumbling into controversy in public, keeps drawing attention, and keeps taking on new projects. It’s genuinely impressive to watch. He’s still only around 25. Whatever he does next is going to be worth watching. For what it’s worth, I’m personally fascinated by this studio-house model he’s running — capital plus know-how, scaling multiple apps globally under one roof. I’d love to build something like that myself one day. But I know I need to put more points on the board first before that’s a credible plan. I’m going to keep publishing case studies like this one — and keep testing this stuff myself, on the front lines, building my own apps in the meantime. Let’s keep at it together. That’s it for this week’s deep dive on Blake Anderson. Thanks for reading all the way through — if anything stood out to you, or you have questions, just reply to this email. I read everything. Reference https://x.com/blakeandersonw https://www.10x.so/ https://www.10x.so/about https://www.10x.so/team https://www.10x.app/ https://www.10x.app/about https://x.com/EvanYadegari https://appmafia.com/ https://il.ly/blog/app-mafia https://techcrunch.com/2026/03/02/myfitnesspal-has-acquired-cal-ai-the-viral-calorie-app-built-by-teens/ https://techcrunch.com/2025/03/16/photo-calorie-app-cal-ai-downloaded-over-a-million-times-was-built-by-two-teenagers/ https://www.cnbc.com/2025/09/06/cal-ai-how-a-teenage-ceo-built-a-fast-growing-calorie-tracking-app.html https://julianivaldy.medium.com/viral-app-playbook-562c728670be https://www.readsocialfiles.com/p/2-founders-90-days-1m-arr https://www.linkedin.com/in/blakeandersonw/

Local AI Is Not Enough Running a model on your own hardware changes where inference happens, but not necessarily what the model knows, how it was trained, or who controls its future. There is an obvious appeal to running AI locally. Your prompts do not need to travel to a remote API. Your files can remain on your own network. You are not dependent on a provider keeping the same pricing, access rules, or product priorities. If the service disappears, your local system can still run. That matters. I have been building my own local AI setup for precisely these reasons. I want an assistant that can work with me without every interaction becoming a request to a distant platform. I want to understand the system I am using, shape it around my own workflow, and keep more control over the data that passes through it. But local execution is only the first layer of the problem. A model running on your machine is not automatically transparent. A model with downloadable weights is not automatically open source. A model with an open license is not automatically governed by the people who depend on it. These ideas are related, but they are not interchangeable. The technical distinction is important because our language shapes what we demand from AI developers, what we audit, and what kind of infrastructure we build next. My local AI experiment In the video that prompted this article, I described the practical tradeoff I keep encountering in local inference. I have a GX10 system (DGX Spark equivalent) and a RTX 5090, and I have been using them to run large local models for agentic work. The goal is not to chat with a model. The goal is to give an assistant enough capability, memory, and tool access to become useful in real workflows. The tradeoff is familiar to anyone who has tried to run larger models locally. More capability usually means more memory pressure and more demanding hardware. Smaller models may respond faster, but they can be less capable on complex reasoning, coding, or multi step tasks. A larger model may produce better work, but at a speed that makes interactive use frustrating. On my 5090 setup, Qwen 2.6 27B Q6 reaches roughly 100-110 tokens per second while on my GX10 produced roughly 10 tokens per second, while on that same GX10 Qwen 3.6 35B A3B reaches 50-70 TPS. Those are measurements from my own environment, not universal benchmarks. They are still useful because they show what the decision feels like in practice. The question is not simply, “Which model is smartest?” The practical question is, “Which model is capable enough, fast enough, affordable enough, and controllable enough for this task?” Current model releases make the hardware range visible. The official Qwen3.6 35B A3B model page describes an open weight model with a default context length of 262,144 tokens and compatibility with local inference frameworks including Transformers, vLLM, SGLang, and KTransformers. Google’s Gemma 4 model card describes a family ranging from models intended for mobile and laptop deployment to 31 billion parameter dense models and a 26 billion parameter mixture of experts model. The direction is clear even without treating vendor benchmarks as universal truth. Local inference is becoming a spectrum rather than a single category. The same broad model family can target a phone, a laptop, a desktop GPU, or a server. The best choice depends on the task and the constraints. Local, cloud, and hybrid are different deployment choices It is tempting to turn local AI into an ideological alternative to cloud AI. I do not think that is technically honest. Cloud systems still offer access to models, tools, and infrastructure that many people cannot reproduce at home. They can be useful when a task needs more capability than the local system can provide. Local systems can be useful when privacy, cost predictability, latency, availability, or customization matters more than maximum capability. For many users, the realistic architecture will be hybrid.

Anthropic is donating another $20 million to Public First Action We're contributing an additional $20 million to Public First Action, bringing our total support to $40 million. Public First Action is a nonpartisan organization that educates the public about AI and works with Republicans, Democrats, and Independents who are serious about putting sensible AI safeguards in place. Both of our donations were made exclusively to support Public First Action’s public education and policy mission, and cannot be used to influence the election of any candidate for federal, state, or local office. Our first donation in February 2026 was made to help promote policies that will maintain meaningful safeguards, sustain America’s AI leadership, and demand transparency from the companies building the most powerful AI models. In the months since, the case for these policies has only gotten stronger. Why now? AI models continue to improve at a rapid pace. Earlier this year, Claude Mythos Preview discovered thousands of high-severity software vulnerabilities, including some in every major operating system and browser. We chose to release it to a limited set of cyber defenders through Project Glasswing to enable trusted actors to find and fix those weaknesses before anyone could exploit them. In the wrong hands, models like this could threaten the critical systems the country relies on, from hospitals to the energy grid. More capable models will bring benefits—like compressing drug discovery, developing treatments for diseases we’ve never been able to treat, accelerating science to deliver a century of progress in a decade, extending human lifespans, and driving the kind of economic growth that lets everyone share in the prosperity—but we need to make sure we are protected against the risks first in order to realize those benefits. Governments need time to build the capacity to capture benefits while containing risks, which is why they need to start now. Policy that meets the moment We’ve long argued that frontier AI companies should be transparent about what their models can do and how they’re managing the risks. We’ve supported newly passed laws in several states that require greater transparency for AI developers. But given how fast the capabilities of the most powerful models are advancing, transparency alone is insufficient. We need policymakers and candidates to put forward measures that mitigate risks. This is the core of what we’ve laid out in Anthropic's Advanced AI Framework, which is the strongest policy proposal from any frontier lab or policymaker to date. Governments should be able to verify companies’ safety claims, enforce safe practices through civil penalties, and ultimately have a way to slow or block the deployment of AI models that pose a serious risk of catastrophic harm. Frontier AI developers should have to test models that pose catastrophic risk, be transparent to the public about their findings, submit them to independent evaluation, and maintain a robust security program. As the pace of AI advances accelerates, the national security stakes of AI are growing, as we outlined in 2028: Two Scenarios for Global AI Leadership. America and our democratic allies have the advantage today, but that lead is tenuous. To ensure continued American AI leadership, we support policy efforts to tighten export controls on advanced chips and semiconductor manufacturing equipment and to curb illicit model access and distillation attacks, so that democracies’ best technologies are not used to advance authoritarian AI. What comes next? These policy frameworks are just a starting point. Our donation to Public First Action is one way in which we’re trying to raise the salience of this urgent policy debate.