Up to $300M Poke Deal Shifts Devin From Tool to AI Colleague
Cognition acquired conversational AI assistant Poke for up to $300 million to integrate personality-driven persistent memory into the Devin coding agent.
Cognition, the AI lab behind the Devin autonomous coding agent, acquired the creator of Poke in a transaction valued between $100 million and $300 million. The purchase of The Interaction Company of California targets the integration of Poke’s conversational character layer into Cognition’s core SWE-1.7 software-engineering model.
Infrastructure Consolidation
Launched in March 2026, Poke operates as a personality-driven agent on iMessage, SMS, Telegram, and WhatsApp. It facilitated over 100 million messages in its first three months and became the first standalone third-party AI agent approved for Apple’s Messages for Business platform.
Despite high adoption serving hundreds of thousands of daily users, high computational costs created scaling challenges for the standalone consumer product. The merger brings Poke under Cognition’s proprietary computing infrastructure to ensure long-term sustainability.
Cognition co-founders Scott Wu and Walden Yan, who were early angel investors in Poke, are driving a rapid consolidation phase. This marks the company’s third major acquisition, following the purchase of post-deployment automation firm TierZero on July 20, 2026, and the integration of Windsurf in 2025.
From Utility to Colleague
As code generation commoditizes across tools like GitHub Copilot and Cursor, AI personality and conversational user experience are emerging as competitive moats. Cognition is actively moving Devin away from a transactional, command-line utility.
The integration of Poke aims to create a persistent engineering colleague. Poke’s architecture maintains a consistent behavioral profile across long sessions and context switches. If you rely on AI agents to manage parallel workstreams, this persistent memory prevents the agent from losing context over multi-day tasks.
Enterprise pilots testing Poke’s character layer demonstrate the material impact of this approach on engagement.
| Metric | Standard AI Assistant | Poke Character Layer |
|---|---|---|
| Interaction Model | Transactional utility | Proactive colleague |
| State Management | Single-session | Multi-session memory |
| 90-Day Retention | Baseline | 40% higher |
Phased Integration
Poke will remain an independent consumer service through the end of 2026. Cognition will begin the full technical integration of the character layer into the Devin platform in early 2027.
Developers building multi-agent systems should evaluate how conversational interfaces and agent memory impact their own deployment metrics. Standard transactional interfaces are proving insufficient for complex, long-running agent workflows where context retention directly dictates user retention.
Get Insanely Good at AI
The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.
Keep Reading
How to Build Reliable Agents With Ai2's Shippy Architecture
Learn how to implement the Shippy agent architecture using distinct personas, versioned skills, and deterministic CLI tooling for reliable AI deployments.
Claude 4 Engineering Edition Solves 48.2% of SWE-bench 2026
Anthropic released Claude 4 Engineering Edition with a 2.5-million-token context window, autonomous IDE integration, and per-resolved-issue billing.
SymptomAI Agentic Interviews Beat Clinician Diagnostic Accuracy
Google Research demonstrated that active, agent-driven symptom interviewing yields significantly higher diagnostic accuracy than human clinicians.
Rakuten Speeds Up Agent Workflows 10x Using Claude Fable 5
Anthropic detailed how Rakuten uses Claude Fable 5 to automate multi-step software engineering tasks, resulting in a 10x speedup for complex workflows.
Gemini-Powered Video Agents Secure $4M Pre-Seed for Fika Jobs
Stockholm startup Fika Jobs raised $4 million to launch a recruitment platform where candidates conduct 10-minute screening interviews with Gemini AI agents.