ChatGPT Is Android, Claude Is iOS

January 26, Anthropic. February 3, OpenAI. Within ten days, both companies officially launched external app integrations built on MCP Apps. Asana, Figma, and Slack now run directly inside AI chatbots.

A year ago, this felt like a novelty. Then OpenClaw hit 150,000 GitHub stars, and the perception shifted - people finally understood what a personal AI agent could be.

Layer on the $300 billion SaaS stock wipeout - traders are calling it the “SaaSpocalypse” - and it becomes clear this isn’t a feature update. It’s a platform transition.

The parallels to the 2008 iOS-Android rivalry are striking. Three signals in particular stand out.

Revenue Models Are Diverging

On January 16, OpenAI officially announced advertising in ChatGPT, targeting free-tier and Go subscribers. Sam Altman said in 2024 that “combining advertising with AI feels uncomfortable.” The fact that they reversed course tells you how urgent the economics have become.

Anthropic, meanwhile, ran a Super Bowl ad declaring “no ads in Claude” - a media play that turned a limited consumer audience into a positioning advantage, aimed squarely at OpenAI.

This isn’t a philosophical difference. It’s structural.

• ChatGPT has 800 million monthly users and the top traffic ranking. That’s a scale where advertising works.
• Claude has a smaller consumer footprint but is approaching 50% market share in B2B.

The 2008 mobile market split along the same lines. Android shipped a free OS funded by advertising. iOS charged a premium for hardware and built a paid app ecosystem. The result: Android took 72% of global market share while iOS captured 65% of app store revenue. Two viable models, completely different economics.

Openness Cuts in Different Directions

The interesting part is that this isn’t a simple “ChatGPT = open, Claude = closed” story.

Anthropic created MCP, the open protocol. Claude’s app integrations run exclusively on the MCP Apps standard - nothing proprietary.

ChatGPT adopted MCP Apps but maintains a separate Apps SDK alongside it. Features like Instant Checkout, file uploads, and modal dialogs are delivered through window.openai - a platform-specific extension layer.

Anthropic opened the protocol. OpenAI is opening the platform.

The mobile era had the same dynamic:

• Android open-sourced the OS, but Google Play Services created a proprietary dependency layer.
• iOS kept the OS closed, but standardized the app ecosystem’s rules early, giving developers a controlled but predictable experience.

What stands out is OpenAI’s trajectory over the past year. A company that was relatively closed has aggressively embraced open-source compatibility, expanded developer documentation, and adopted the MCP standard. Compared to a year ago, the urgency to secure an ecosystem is unmistakable.

The Existing Software Market Is Actually Shaking

This week’s SaaS stock collapse is the market beginning to price in this platform shift.

• ServiceNow is down 28% year-to-date, Salesforce down 26%, Intuit down 34%.
• Jefferies traders coined the term “SaaSpocalypse.”
• A single legal automation announcement from Anthropic sent Thomson Reuters down 16% and LegalZoom down 20%.

SaaStr’s Jason Lemkin points out that “the real cause isn’t AI - it’s growth rates declining every quarter since 2021.” He’s right. But the emergence of AI app stores gave the market a reason to reprice. CIOs are already freezing new SaaS purchases and redirecting budgets toward AI infrastructure.

This echoes 2008 - 2012, when PC software companies first started losing customers to mobile apps. The displacement was gradual, then sudden.

What Should We Build on Top of This

In the web era, the browser was the computing entry point. In the mobile era, the OS took that role. In the AI era, the abstraction rises one more layer. Users won’t search for and install apps themselves - the chatbot understands context and summons the right tools automatically. The entry point shifts from human to AI.

But this chatbot isn’t hardware you own. It’s closer to a virtual computing machine that operates on your behalf. Projects like OpenClaw and Claude Code are gaining traction precisely because of this: the desire to run your data, on your device, with a model you choose.

What every service built on this new form factor must ultimately prove is simple: can it deliver value that exceeds the fixed cost of the AI consuming it? People spend thousands on hardware without hesitation but balk at paying for services. That’s the same psychology from early mobile - “why would I pay for an app?” - and we’re in the same transitional period.

Web to mobile. Mobile to AI. Every platform shift brings an app war on top of it, and this one has just begun.