A Race Every AI Company Is Running

The timeline tells the story. November 2025: Peter Steinberger, the Austrian developer who built PSPDFKit into a tool used by Apple and Dropbox, open-sources a side project called Clawdbot — a messaging-based AI agent you control over WhatsApp. January 30, 2026: after rebrands to Moltbot and then OpenClaw, the project crosses 100,000 GitHub stars in a single week. February 15: Steinberger joins OpenAI. Sam Altman calls him “a genius” who will “drive the next generation of personal agents.” Steinberger’s condition: OpenClaw stays open-source, moving to a foundation. March 3: OpenClaw surpasses React and Linux as the most-starred project on GitHub — 250,000 stars, two million monthly active users. March 20: Anthropic ships Claude Channels.

The pattern is simple. The AI people can reach without switching tools or waiting for someone else to operate it — that’s the AI that gets used. And the AI that gets used compounds its advantage through adoption, feedback, and trust. OpenClaw proved it: 92% retention. Once people could message an agent from their phone, they didn’t go back to terminals.

Claude Channels is Anthropic’s answer. Rather than building another standalone app, Anthropic embedded Claude into Discord and Telegram — the platforms teams already depend on. Because Channels connects to your actual codebase via MCP, it’s not a chatbot. It’s an agent that reads, writes, and modifies real code on your machine while you’re away from your desk.

The tools that win will be the ones that disappear into existing workflows.