BRIEF
Meta's Muse agent hit 5 million US downloads in 22 days, faster than ChatGPT, Grok, or Claude.
Anthropic and OpenAI built their pitch on building God. Meta and Google are building products.
If Big Tech peels customers away, the frontier-lab IPOs (OpenAI's next year) lose their story and their buyers.
Concept Artist: Alan Zhou / Creative Studio: Sum of Parts for Zero Knowledge🤦♂️
So who needs a god?
Good morning,
Anthropic and OpenAI pitched a god.
Meta shipped an app that feels like a labubu.
Guess which one people downloaded.
Grumping,
Austin Campbell
Teaching DeFi at NYU Stern
Bridging TradFi at Zero Knowledge
P.S. The Zero Knowledge team works hard to produce this for you.
If you like it, share it with your colleagues.
🏹 BACK UP
WTF ARE FRONTIER LABS?
Frontier labs are the handful of companies, chiefly OpenAI and Anthropic, that build the largest and most capable AI models and sell them as the path to something close to general intelligence.
WTF IS MUSE?
Muse is Meta's personal AI agent app, launched September 8 in the US and Canada. It is a product wrapped around a model, not a model sold as a destiny.
22 DAYS
⏳ BREAKDOWN
In five weeks, a launch turned into a test of the frontier-lab thesis.
SEP 8
Meta launches Muse in the US and Canada.
SEP 30
Muse passes 5 million US downloads and sits at No. 1 on the App Store.
OCT 5
Similarweb counts 2.18 million daily active users.
OCT 12
Sensor Tower data shows about 73% of new users return the next day, roughly double ChatGPT. ChatGPT overtakes Muse between days eight and nine.
NEXT
Anthropic's IPO is reportedly planned for November. OpenAI's follows next year.
ZERO INSIGHTS : THE LONG READ
AI Adoption and the Scurvy Puzzle
"James Lind — Conqueror of Scurvy" (Robert A. Thom, 1959) — CC BY-SA 4.0
The history of technology shows that even the most powerful innovations face a long path to full adoption. Lemons, a proven cure for scurvy, took decades to reach key users like the British Navy.
Despite that history, many AI leaders seem to assume adoption will take care of itself. Sam Altman called it “inevitable.”
It’s not.
🎭 BUSTED
THEY SAY
Frontier models will cure cancer.
BUT ACTUALLY
Tools that are genuinely useful in advanced fields take more than throwing a language model at the problem. Mathematician Terence Tao's new "Math 2.0" talk argues humans must stay in control of, and central to, advanced research.
~0.7%
💰 BAG HOLDERS
WINNERS
Meta, which turns a model into a product people use daily.
Google, which has the distribution and the business model to subsidize costs.
Consumers, who get lighter, cheaper tools built into things they already use.
LOSERS
Data center investors, who financed capacity for demand that may not arrive.
OpenAI and Anthropic, whose IPOs depend on the frontier story holding.
Oracle, a counterparty on the data center buildout.
ZERO OUT
Magic is a great pitch and a bad business model.
BEST CASE
The techno-utopians at OpenAI and Anthropic realize they have to compete like businesses, and start shipping products people want.
WORST CASE
OpenAI's IPO pairs a section on how AI could kill us all with dismal financials, and the whole thing melts down.
The ideas behind these labs produced remarkable data-processing technology. Mistaking that for actual magic is the risk.
ZERO INSIDER
Austin joined News Nation's Decision Desk '26 to discuss the Ohio governor's race, where a new ad featuring Democrat Amy Acton's mother calling her daughter a "liar" has stirred outrage. Kalshi's market on screen had Acton at 74% against Vivek Ramaswamy's 26%.
David presented his upcoming Tornado Cash documentary in a private preview for backers and distributors in Manhattan. The six-hour deep dive into encryption history and law will arrive some time before Roman Storm’s re-trial, now scheduled for April. You can read an excerpt of the script here.