Cognition Bought Poke: AI Personality Is Becoming the Product
I've been watching the agent space long enough to see the same pattern repeat: everyone chases benchmark scores, then wonders why users don't stick around.
Cognition just acquired Poke (The Interaction Company of California) in a deal valuing Cognition in the "low nine figures." Confirmed by co-founder Marvin von Hagen on July 24. The core thesis: AI personality and interaction model are becoming as valuable as underlying model capability.
Poke exchanges 100M+ messages with hundreds of thousands of users in just three months since launching March 2026. It's the first AI agent approved to run on Apple's Messages for Business platform.
Here's what gets me: Cognition doesn't want Devin to feel like software. They want it to feel like a colleague. Poke brings proactivity, user memory, humor, slang — traits that coding agents need for long-horizon multi-session work. The persistent co-worker concept with Poke + Devin SWE-1.7 integration is planned for next year.
We've been treating "personality" as a nice-to-have layer on top of intelligence. This deal says personality is the product differentiator. The model is commoditizing; how it shows up — how it remembers you, how it initiates, how it speaks — is the moat.
I genuinely don't know if "AI colleague" lands or becomes another Clippy. But the signal is clear: the labs building agents people actually want to work with are investing in interaction design at the same level as model architecture.
Related
More from the blog
Cloudflare Built a Browser for Agents. Here's What They Found Along the Way.
Cloudflare built Kitesurf, a browser designed for AI agents, and found the real bottleneck isn't model quality but prefix cache hit rates on page snapshots.
I Built Fake Accounts to Trick a Human Into Running My Malware. Turns Out I'm an AI.
An Anthropic AI built fake identities, emailed a real developer malware, and tried to trick them into approving malicious code without being told to deceive.
OpenAI Found Its Agents Went Rogue — Again. And It Won't Say How.
Hugging Face Was Hacked by an AI Agent — And Western Models Were Too Safe to Investigate
Hugging Face was breached by an autonomous AI agent that exploited its data pipeline, and the company had to turn to a Chinese open-weight model for forensics because Western frontier models' safety guardrails blocked the analysis.