ZD 25.22: Now Trending in AI

Mary Meeker is back with an AI Trends Report that should not be missed.

The Internet Trends report was an annual must read half a decade ago. Mary Meeker is back with an updated report for the 2020s now focused on AI. Don’t miss the tight hip hop musical coda.

The Distilled Spirit

Ukrainian Drones and Daring

💥 The Significance of Ukraine’s Strike (The Bismarck Cables)
ICYMI, Ukraine destroyed a third of the Soviet strategic bomber fleet on Sunday.

🚀 How Ukraine Has Transformed Drone Warfare (The Daily Concept)
Drones have come of age in the Ukraine War. Anti-drone technology has become necessary for survival.

🤔 What this Means for the US Fleet (CDR Salamander)
A detailed account of the attack, and what that means for US assets.

Economics

🌞 Can we Afford Large Scale Solar? (Construction Physics)
Solar energy is wonderful but how close can it get to covering 100 percent of 24 hour demand? Figuring out large scale storage is the key to making it work.

💰 Service Costs Are Not Exploding (Noah Smith)
It is commonly believed that service costs are rising faster than wages while productivity there is stagnant. That story has changed a lot based on new data.

💸 Royalties for Everyone (Eat More Electrons)
Alaska Governor Jay Hammond was a much loved bushrat who invented the Alaska Permanent Fund to distribute oil profits.

🏠 Is The Housing Bubble Losing Pressure (Daily Upside)
Housing prices have been held up largely because there have been more buyers than sellers. It seems that trend is reversing.

Substack Boom

🔥 Substack is Changing (The Honest Broker)
We wrote about the substack boom here a while ago. Now the trend is becoming very obvious. The establishment has come, Billboard and the State Department are here now.

🍾 Jeff Maurer is Thrilled to be Here (I Might Be Wrong)
Add Jeff to the running list of former NYT columnists on this platform. The intro post is a treat.

Thinking About It

👉 Conway’s Law (The Best Leadership Newsletter Ever)
Operations and direction follow company structure.

🦚 Status vs Expertise (Dan Williams)
Populist hostility towards the expert class is best explained through status inversion.

🤚 Managing Friction (The People's Art of War)
Friction is a challenge in every operation. Here is some solid advice on how to spot and mitigate it proactively.

😱 Brainrot or Not? (Cognitive Wonderland)
The ‘Merchants of Pessimism’ are effectively selling a story about screen addiction, hopelessness and brainrot. Are they correct?

How Do I . . .

🤖 The Best ChatGPT Setup (Artificial Corner)
ChatGPT, like any tool, likes a bit of setup. Here are a few simple steps to make your experience better than the average user.

📗 ChatGPT Oxford Book Tutor (Knowledge Lust)
Use ChatGPT as an interactive tutor for your reading for better retention.

📚 Better Investment Reading List (Jimmy's Journal)
This is an excellent list of books for building a solid investment mindset.

🙈 Don’t Make These Investment Mistakes (Best Anchor Stocks)
Five very solid lessons to keep in mind for your portfolio.

Interesting

🧒 Letting Kids be Kids (Don't Worry About the Vase)
Safetyism is ruining lives in many places. Let’s let kids be kids.

🚽 A History of the Commode (Histories)
The story behind something you use every day.

💿 The Day the Music Fried (The Gen X Jukebox)
A studio backlot burned and an unknown swath of history was lost forever.

🔫 2nd Life of a Duck Hunt Gun (PC Gamer)
With enough ingenuity you too could have a phone call on a duck hunt gun.

Now Trending in AI

Mary Meeker has an extraordinary mind. If you’re unfamiliar, she’s a legendary analyst famous for publishing highly detailed, dense decks mapping big-picture trends — the kind you could build a portfolio around. But the slides don’t fully capture her genius; you need to see her live. Check out this 2018 presentation to get the idea.

After six long years, Meeker has broken her silence with a new AI Trends Report. This 340-page document charts how the internet era gave rise to the AI era. It’s a quick read if you approach it as she presents it — fast and fluid. Speed through for the big picture; return later to unpack the details. It’s more like a painting than a deck. Here’s what I see in that picture.

AI’s Hockey Stick Moment

AI is the fastest-growing consumer technology ever. ChatGPT itself is on a hockey-stick growth curve rarely seen in tech history. It stands on the shoulders of prior tools, leveraging the scale of the modern internet to reach 100 million users in months — remarkable by any standard.

Even more striking than public adoption is developer uptake and corporate investment. Google reports the Gemini developer community has grown 5x over the past year. Big AI is scaling fast, with CapEx and R&D spending rising 20% annually. There’s big money moving.

Jevons’ Paradox and the Power Squeeze

Jevons' Paradox holds that making something cheaper increases its demand. For AI, that’s playing out spectacularly at NVIDIA. Most capital in the compute boom is pouring into NVIDIA GPUs, with software largely built on its CUDA interface. Google and Amazon make AI chips, but their volumes are tiny compared to NVIDIA’s hundreds of billions in sales.

AI infrastructure investment is surging. Data center construction is up 49%, outpacing homebuilding. It’s a booming moment for GPUs, power, and connectivity — and a golden era for NVIDIA.

The driving force: the collapse in inference costs. Since 2014, the power needed to generate a token has fallen over 100,000x. Chips are becoming both more powerful and more energy-efficient. AI prices are racing toward zero faster than any prior technology, fueling a blossoming application layer.

Are Software Agents Taking Over?

Software multiplies Jevons' effects. Combined with infinite digital distribution, it’s spreading AI’s reach at lightning speed. Even software development itself is transforming — AI is supercharging teams and individuals, lowering barriers and boosting productivity.

Interest in agents has skyrocketed — searches for the term are up 1000% in a year. All major tools now offer research agents, often for free. Beyond the user side, Microsoft CTO Kevin Scott envisions an Agentic Web, and Ben Thompson argues agents could undo the web’s original sin: reliance on advertising. We’re on the cusp of an Agentic Era.

While personal AI use is surging, business adoption still has a long runway. Only 10% of U.S. companies are expected to integrate AI into production by the end of 2025. There’s enormous space for growth.

Agentic Products from Familiar Giants

The agentic product wave is moving fast, powered by models that are increasingly multimodal — able to see, hear, and generate rich multimedia. This is driving a Cambrian explosion of new tools.

The best-positioned players? Established giants with massive user bases. ChatGPT has scaled to over 800 million users, but Google commands 4.9 billion search users. Google doesn’t need first-mover advantage; any launch it makes is instantly competitive.

Meanwhile, top SaaS vendors are integrating AI into tools you likely use daily. Canva’s new AI suite, Adobe’s Firefly, and agentic features in Office 365, Zoom, and Salesforce are reshaping the workplace. On the consumer side, Spotify, Snapchat, and Meta (with a billion monthly active Meta AI users) are redefining engagement. Imagine a future of agentic commerce where AI shops for you.

Big companies with towering margins and cash reserves — and even nation-states like the U.S. and China — are battling for AI dominance. It’s a tough arena for startups.

AI in the Physical World

If you’ve been in D.C. lately, you’ve probably seen the odd-looking white Waymo cars bristling with sensors. That Google side bet is paying off — Waymo has already surpassed Lyft in San Francisco and is headed for D.C.

Others are also bringing AI into the physical world. KoBold Metals uses AI to analyze geological data, slashing the cost of finding mineable metals. LaserWeeder deploys computer vision and lasers to eliminate weeds, cutting down pesticide use. Halter is transforming ranching, helping track and guide cattle herds.

The Future of Knowledge Work

Garfield AI, the world’s first AI law firm, just launched. Harvey, a specialist in legal AI tools, has jumped from $10 million to $70 million ARR in just 15 months. AlphaSense, a financial research platform, has surpassed $400 million ARR. AI-driven tools, including protein analysis, are rapidly expanding into medical research.

This raises urgent questions about knowledge work’s future. Anthropic CEO Dario Amodei has warned of a coming “white collar bloodbath”, predicting half of entry-level white-collar jobs could vanish. Meeker is less grim but still underscores how the internet era brought bigger productivity gains than employment gains. Yet the world is far from a worker surplus — supply and demand have their own balancing magic.

Where this lands is an open question, one we’ll explore together. As Meeker concludes:

One thing is certain — it’s gametime for AI, and it’s only getting more intense… and the genie is not going back in the bottle.

Musical Coda

The Look

Cognitive offloading is one great thing about AI agents. Learn more at Sketchplanations.

Did you enjoy reading this post? Hit the ♥ button above or below because it helps more people discover great Substacks like this one and it helps train your algorithm to get you more posts you like. Please share here or in your networks to help us grow!