Consumer AI
a16z just dropped their third consumer AI report. I read every edition. Every time the picture is completely different.
3 years ago the line was clean: AI products over here, everything else over there. That line is gone. CapCut: 736M MAU, most popular features are AI. Notion's paid AI feature usage went from 20% to 50% in a year. AI stopped being a product category. It's now the base layer inside every product.
5 things that stood out:
1/ The "default AI" race is just getting started
ChatGPT sits at 900M WAU, 8x more paid subscribers than Claude and 4x more than Gemini. Dominance is obvious. But competitors are growing faster for the first time: Claude +200% YoY, Gemini +258%. And 1 in 5 ChatGPT users already uses Gemini in parallel. There's leadership. There's no loyalty.
Platforms are diverging strategically. ChatGPT is building a superapp (shopping, bookings, ads, "Sign in with ChatGPT"). Claude is going after power users (dev infrastructure, financial terminals, MCP connectors). This isn't a search-style winner-takes-all market. It looks more like iOS vs Android, 2 ecosystems, both can survive.
The real lock-in is context. The more AI knows about you, the better it works, the harder it is to leave. Whoever becomes the operating environment, not the chat window — wins.
2/ One product, one AI task no longer works
Platforms are expanding and absorbing everything that used to be standalone. In 2023, 7 of 9 creator tools in the a16z ranking were image generators. Now it's 3. Midjourney fell from the top 10 to #46. ChatGPT and Gemini shipped their own image models, and for most users that's enough.
If your product does exactly one thing, the platform will add it as a feature. Two types survive: the superapp and the deep vertical with proprietary data. Everything in between is at risk. Image generators are already mostly dead. Video and music are holding on (Suno, ElevenLabs) but platforms are coming.
3/ From content generation to action generation
Platforms ate text, images, soon video. But content is passive output. The next layer is AI that doesn't generate. It acts: books appointments, fills forms, coordinates workflows, sends things on your behalf.
OpenClaw, an open-source agent by a solo developer in Austria, hit 68,000 GitHub stars in weeks, surpassing React and Linux. Acquired by OpenAI. Meta bought agent platform Manus for ~$2B. Big companies aren't building agents. They're acquiring whoever already has traction.
Agents are still far from mainstream. But once users connect AI to their calendar, email, and CRM, switching costs become enormous. Connectors are the new lock-in.
4/ AI is becoming invisible
AI is no longer a site you visit. It's embedding into tools you already live in.
Claude Code hit $1B ARR in 6 months. OpenAI Codex (Mac desktop app): 2M WAU, growing 25% weekly. Claude for Excel and PowerPoint. Gemini inside Docs and Gmail. None of these show up properly in standard rankings, because rankings count web traffic and mobile MAU, and these products live in the terminal and the desktop, inside other interfaces.
a16z admits their own methodology is breaking. The fastest-growing AI products are invisible to the metrics their report is built on.
A $1B ARR product with no landing page and no marketing team. It lives in a terminal. This breaks every old rule about distribution.
5/ Russia is the third AI pole
The global AI market has split into 3 ecosystems that barely overlap. The West (ChatGPT, Claude, Gemini). China (Doubao, Kimi). And Russia, for the first time explicitly called out as a separate ecosystem in the a16z report.
Yandex Browser with Alice: 71M MAU, top 10 mobile AI products globally. Sber's GigaChat made the web ranking. DeepSeek is the only product operating across all 3 worlds: China (33.5%), Russia (7.1%), US (6.6%).
For anyone building AI in Russia: the competition isn't OpenAI. It's Yandex and Sber, who are already in the global top 10.
TL;DR: AI went through 4 phases in 3 years. Standalone product, then feature inside existing products, then platforms absorbed standalones, now it's leaving the browser and becoming invisible. Visibility dropping, impact growing.
P.S. By AI adoption per capita, Singapore leads. The US is 20th. Building AI and using AI are different skills.
Stay tuned
🥷🥷🥷