Cognition acquires AI texting assistant Poke for low nine figures to enhance Devin's conversational UX.
For AI coding agents, raw generation capability is plateauing, making developer experience (DX) the new moat. By acquiring Poke, Cognition aims to reduce the friction of steering complex code generations through a more natural, state-aware conversational interface. This signals a shift from treating LLMs as command-line utilities to collaborative pair programmers.
Cognition, the startup behind the autonomous AI software engineer Devin, has acquired consumer AI texting assistant Poke in a deal valued in the low nine figures. Poke is known for its highly engaging, friend-like conversational interface. The acquisition aims to integrate Poke's interaction model and conversational UX directly into Devin.
Technical Details While foundation models dictate raw reasoning and generation capabilities, the orchestration layer—how the agent interprets intents, asks for clarification, and maintains context across a session—is highly dependent on the UX. Poke's architecture likely excels in stateful, multi-turn dialogue management and personality alignment. By porting these interaction paradigms to Devin, Cognition is likely building a more robust steering mechanism. Instead of users writing rigid, prompt-engineered specifications, they can interact with Devin through iterative, natural language refinement, much like chatting with a senior engineer.
Why It Matters From an engineering perspective, this acquisition highlights a critical evolution in the AI agent space: raw model intelligence is no longer a sufficient moat. As frontier models from OpenAI, Anthropic, and Google converge in coding benchmarks, the differentiator becomes Developer Experience (DX). If an AI agent cannot effectively communicate its plan, ask the right clarifying questions, or gracefully handle course corrections from the user, its underlying intelligence is bottlenecked. Cognition is betting that an empathetic, highly responsive interaction model will reduce friction in human-in-the-loop workflows, increasing the overall success rate of complex coding tasks.
What to Watch Next Keep an eye on how Devin's interface evolves in the coming months. We should expect to see new features like proactive status updates, more natural clarification prompts, and a shift away from terminal-style outputs toward conversational collaboration. Additionally, watch for other AI agent platforms (like GitHub Copilot Workspace or Magic.dev) to make similar investments in UX and personality-driven interaction models to remain competitive.