🌟 Editor's Note
Recapping the AI landscape from 07/21/25 - 07/28/25.
🎇✅ Welcoming Thoughts
Welcome to the 3rd edition of NoahonAI.
What’s included: company moves, a weekly winner, AI industry impacts, practical use cases, and more.
We’re back! A test email triggered an issue that briefly locked my account this morning. Support was slow so I messaged the Beehiiv team on Linkedin and they were quick to help.
If you saw the huge data leak from social media app Tea this week, I’d venture to guess it was caused by poor vibe coding. The giveaway: No cybersecurity!
Last week I mentioned including a section on the non-NVIDIA5 companies, but I included a ton of info about them in my Basics & Buzzwords blog so we’ll keep it there and keep it updated.
Interview this week is from a talented founder in the music space. Great stuff if you’re interested in how AI can be used in running your own company.
Very much a solid week across the board for the NVIDIA5, no major blunders, and a lot of positive numbers coming out.
The other thing the NVIDIA5 have in common? They always need more cash!
I gained access to GPT Agent, very cool! More on that next.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: OpenAI // ChatGPT
GPT-5 is coming soon; Expected to be released in early August and include improvements to responses, memory, and speed. GPT-5 will handle multi‑step tasks on its own, text + code, and include versions that can fit into everyday apps and devices. I’m ready!
I tried out GPT Agent (just got access this week) and have some thoughts. The software isn’t all the way there yet, but you can see where it’s going. I asked it to complete a few tasks on a to-do list I had in docs. It found the list, and it completed a few basic research based tasks on its own, but it kept moving the date for a flight I asked it to book. The UX of watching it work is pretty cool, and It will pause to prompt input for tasks like logins. I’d recommend trying it out yourself to get a base idea of how it will be used once its 100% there. Agentic AI is the next big leap and you’ll be able to see why.
ChatGPT is now handling a whopping 2.5 Billion prompts per day and has more than doubled in usage over the past 8 months. The only negative surrounding OpenAI right now is the consistent need for more cash. It may need to consider a complete move to for-profit in order to keep up financially.
From Top to Bottom: Open AI, Google Gemini, xAI, Meta AI, Anthropic, NVIDIA.
⬇️ The Rest of the Field
Who’s moving, who’s stalling, and who’s climbing: Ordered by production this week.
🟣 Google // Gemini
Gemini in Watches
Google has deployed Gemini across smartwatches, includes devices from Samsung, Pixel Watch, and more. Keep an eye on the AI Wearables market.
450 Million Monthly Users
Q2 earnings call showed the Gemini app reached 450M monthly users, with daily requests surging 50% since Q1. Across all apps, Google has doubled its monthly token usage from earlier this year.
Success in Mathematics
Gemini Deep Think (in addition to an OpenAI model) solved 5/6 International Math Olympiad (IMO) problems within the 4.5-hour time limit, earning a gold medal score. Currently on par with the world’s smartest high schoolers.
🟠 Anthropic // Claude
Targeting $150B Valuation
After some concerns about financial issues, Anthropic announced it’s in early negotiations to raise $3-5B and double its previous $61B valuation in the process.
Strong Enterprise Revenue
The subscription-based model is paying off: 80% of its revenue now comes from enterprise users. Its ARR has grown to $4B since the start of 2025 (4x growth).
Continued Regulation Push
Anthropic formally committed to the EU’s AI Code of Practice, a voluntary framework requiring transparency, risk testing, and disclosure. Companies like Meta refused to sign.
🔵 Meta // Meta AI
Investors Remaining Optimistic
Meta is projected to report approximately $44.8 billion in Q2 revenue (up ~15% YoY). Analysts remain bullish with AI spending planned at ~$70 billion for 2025.
Teaming up with Amazon
In a rare collab, Meta and AWS unveiled a joint AI startup accelerator on July 21, aiming to foster AI innovation globally.
GPT Co-Creator → Chief Meta Scientist
Zuck announced that Shengjia Zhao, co‑creator of ChatGPT and former OpenAI lead, will serve as chief scientist of the newly formed Meta Superintelligence Labs. Meta certainly has the people in place.
🔴 xAI // Grok
Partnership with Kalshi (Prediction Market / Gambling)
xAI partnered with Kalshi, a U.S. regulated trading platform, to integrate Grok AI into prediction‑market interfaces. Grok will be embedded into Kalshi for AI insights and summarizations.
Pursuing $12B in debt financing
Weeks after closing a $10B debt/equity raise, xAI is in negotiations with Valor Equity Partners to secure up to $12B in high‑yield debt. The money is earmarked for NVIDIA chips for new data centers.
Video Gen on the Horizon
xAI is preparing a major Grok update introducing infinite image and video generation directly into the X app. Curious how this will compare to SORA (Open AI - OK) and Veo3 (Gemini - Excellent).
⚪️ NVIDIA
Chip Smuggling
People used to smuggle gold, or jewels, or drugs across country lines… now they smuggle something a bit more valuable: NVIDIA chips. Over $1B worth of high end chips made their way into China during the 3 month US export ban.
Chip-Location-Verification
U.S. regulators are weighing embedded location verification features for advanced AI chips to enforce export compliance and trace unauthorized movement.
Business Still Booming
Internal reports confirm NVIDIA is selling every AI chip it can manufacture, and with NVIDIA5 companies ramping up data centers across the US, demand should continue.
📢 Impact Industries 🏥
Marketing // Vogue AI Model
A two-page Guess advertisement featured in Vogue's August 2025 issue showcases a fully AI-generated model created by the AI agency Seraphinne Vallora. The ad clearly states “Produced by Seraphinne Vallora on AI.” While it wasn’t part of the editorial shoots, its inclusion sparked public backlash over authenticity, representation, and the impact on human models.
Medical // AI Agent Hospital
Tsinghua University unveiled the world’s first AI Agent Hospital, staffed by 42 AI doctor agents across 21 specialties. These virtual doctors autonomously diagnose and plan treatment, having processed 10,000 simulated cases with approximately 93% diagnostic accuracy, while operating within a closed-loop virtual hospital environment.
👨💻 Practical Use Case: Basic AI Workflows
Difficulty: Mid-Level
AI workflows are evolving fast and I tend to look at them in three levels of complexity:
Hybrid workflows: Connecting AI tools and regular apps while including some simple custom code using AI. Example: Taking CRM data, running it through GPT with custom prompts, and creating personalized emails automatically.
Custom workflows.: These are built from scratch with code and are tailored specifically to a business process. Example: An AI agent that handles lead qualification end-to-end, from scraping data to booking calls.
At the end of the day, a workflow is simply the way we get things done. It can be as straightforward as scheduling a meeting or as complex as managing an entire customer onboarding process. In the 1980s, almost every workflow was manual. The internet automated some. The SaaS boom of the 2010s pushed automation even further. Now we’re at the next shift: moving from automated workflows to AI-driven, agentic workflows where the software doesn’t just follow steps, but makes decisions and adapts in real time.
The best way to learn the basics if you aren’t already familiar is by using a tool like Zapier and browsing popular AI automations. Learn how others are using AI and try one out yourself. Good luck!
🎙 Weekly Interview: Will Matz
Will Matz
🏠 Background: Will attended The Ohio State University and was recently named to the Forbes Columbus 30 under 30.
💼 Work: Co-Founder & CEO of Superfan, a social app for passionate music fans.
🚀 Quote: “Go build stuff. Just like find problems or opportunities and stuff and just, you know, build stuff. It's the best time in history to do it. It's easier than ever.”
Noah Weisblat: I’m here with Will Matz, Co-Founder and CEO of Superfan, Will how are you doing?
Will Matz: Good, thanks for having me.
Noah Weisblat: Tell me about Superfan. Why should someone use it, and are you using AI in the app or in development?
Will Matz: Superfan started as a passion project and I’ve been full-time for about a year and a half. It’s basically a weekly Spotify Wrapped for you and your friends. You sign up, connect your music, and every week we wrap up what you listened to and create a social experience around it. We’ve had early users stick with us for years as we’ve built it out. Long-term, we want to create engaging ways for fans to connect directly with artists.
We do use AI. In the product, every user gets an “AI Vibe of the Week,” kind of like a music horoscope. People love them and share them, which has driven growth. On the dev side, we use tools like Cursor and Claude Code heavily to build faster.
Noah Weisblat: Outside of Superfan, what AI tools are in your stack right now?
Will Matz: Mostly Claude Code, Cursor, and ChatGPT. I bounce between them depending on the task. Our codebase is big, and I’ve found Cursor indexes and navigates it better than Claude Code right now, but that might be because I’ve spent more time optimizing Cursor. Onboarding to Cursor was honestly magical compared to VS Code—it just felt so much more powerful.
Noah Weisblat: With AI in general, have you looked into models like Gemini or Grok?
Will Matz: Not deeply. There’s been so much information overload that unless something really fits my workflow, I don’t chase every new model. I’ve heard good things about Grok’s latest updates, but I’m already set up with tools that work. At this point, I just want the AI to work,I don’t care about context window sizes or benchmarks.
Noah Weisblat: How much more efficient would you say AI has made you?
Will Matz: Probably 2–3x. But it’s not just speed—it’s quality. If you use these tools well, you can ship higher-quality code with less effort. Back in 2018, I’d be banging my head against a bug for hours, digging through Stack Overflow and GitHub issues. Now I don’t get stuck like that anymore. That’s the biggest win: removing that deep frustration and blocker time.
Noah Weisblat: What’s your workflow like when you’re building with AI?
Will Matz: I create really detailed specs before making changes so the AI can execute them well. Claude tends to infer specs better, while Cursor needs more prompting. It’s not magic, but the key is writing good directives, letting it rip, and catching errors early if the AI goes off-track.
Noah Weisblat: How do you use AI personally outside of Superfan?
Will Matz: Mostly as a sounding board. One limitation I see across all models is taste—they don’t understand nuance or what makes something truly engaging. For example, our “Vibe of the Week” has a specific tone, and AI often misses it. Same with Gen Z language; it comes off cringy.
Where it shines is starting from a blank slate. We recently used it to mock up a new admin dashboard from scratch, and it nailed ideas we hadn’t even thought of. That was one of the coolest use cases we’ve had.
Noah Weisblat: If you were advising a high school senior or new college student on leveraging AI, what would you tell them?
Will Matz: Build stuff. It doesn’t have to be a startup—just create things. Software, design, whatever. It’s the easiest time in history to do it. Stay on top of tools and use them to bring ideas to life.
Noah Weisblat: Where do you see Superfan in the next few years?
Will Matz: We have ambitious plans. Right now, we’re focused on building engaging fan experiences. The natural next step is bringing artists into the mix, but we’re waiting until the product is ready. AI opens doors for experiences that weren’t possible before, and that’s what excites us most.
Noah Weisblat: Where do you see AI in the next 5–10 years?
Will Matz: Right now, it’s an amazing window to build software. Small teams with coding chops are going to ship a ton of cool stuff and make money, but 95% of those companies will eventually go to zero because AI is moving so fast. The cost of building basic B2B SaaS will drop to almost nothing. That’s why we’re excited about consumer experiences—they feel more durable and meaningful long-term.
Noah Weisblat: Before I let you go, top three artists?
Will Matz: Recently saw Perfume Genius in Cleveland, one of the most engaging shows I’ve ever been to. Huge fan of Richie Mitch and the Coal Miners, they were early Superfan supporters and make incredible indie folk. And Caamp, the Ohio boys. We’ve been fans for years; they’re great musicians and great people.
Noah Weisblat: Excellent. Will, I appreciate your time.
Will Matz: You got it.
🎶 Startup Spotlight
Superfan: Where artists find and activate their top fans.
The Problem: Artists leave billions on the table because they don’t own their fan relationships. Superfans drive the majority of streaming, merch, and ticket revenue, but right now, they’re invisible. Artists don’t know who’s listening to their music or sitting at their shows.
The Solution: We give artists a direct channel to their most valuable listeners. Our fan app has the most comprehensive dataset on gen z fans anywhere, and we’ll use this foundation to help artists find and activate their fans.
The Market: There are 10M+ artists and hundreds of millions of music superfans globally, and they’re underserved by generic platforms that treat all fans the same.
Why Now? AI makes personalization scalable, and the music industry’s top priority is clear: understand and monetize superfans.
The Backstory: We’re two friends who have scaled our music social app to 400K users without paid marketing. Now we’re building the rails to help artists find, activate, and earn from their fans at scale. Others build silos, we’re building the network.
Contact: [email protected]
“It’s not likely you’ll lose a job to AI. You’re going to lose the job to somebody who uses AI”
- Jensen Huang | NVIDIA CEO
If you’re vibe coding, please make sure you don’t store data in public buckets! Till next time,