Cognition's Nine-Figure Bet on Personality for AI Agents

The Deal That Values Conversation

Cognition, the startup behind coding assistant Devin, has acquired The Interaction Company of California in a transaction valued in the low nine figures. The target? Poke, an AI assistant that exchanges messages with users as if it were a friend, complete with slang, humor, and casual tone. Over the past three months alone, Poke users have sent more than 100 million messages through the platform, which operates across iMessage, SMS, Telegram, and WhatsApp in select markets.

The acquisition reflects a growing conviction among AI builders: the way an agent communicates can be as strategically important as the model underneath. Cognition wants to make Devin feel less like a tool and more like a colleague you'd actually want to work with. According to Marvin von Hagen, co-founder of The Interaction Company, the analogy is straightforward. People prefer coworkers with personality over those who act like robots. Even software engineers crack jokes. That dynamic, he argues, should extend to AI agents that write code.

What Poke Brings to the Table

Launched in March of this year, Poke carved out a niche by letting users engage with AI through their preferred messaging apps. People have used it for tasks spanning travel planning, health tracking, financial queries, scheduling, and education. Productivity workflows like managing emails, setting reminders, and organizing to-do lists emerged as the most common use cases, according to von Hagen.

What set Poke apart from other assistants was not the breadth of tasks it could handle but the style in which it handled them. Instead of delivering terse, functional responses, Poke adopted a conversational tone that mirrored how friends text one another. That personality made the assistant feel approachable and, for many users, more enjoyable to interact with over time.

In June, Poke became the first AI agent approved to run on Apple's Messages for Business platform, a standardized channel that companies use to communicate with customers inside the Messages app. That approval gave Poke a foothold in enterprise communications, even as the underlying business struggled with profitability. Von Hagen acknowledged that running Poke had been expensive, making it difficult to generate sustainable margins despite hundreds of thousands of active users.

Why Personality Matters for Coding Agents

Cognition's interest in Poke centers on the belief that personality can improve collaboration between humans and AI agents, especially in technical domains. Devin, which assists developers by handling pull requests and other engineering tasks, currently operates in a more transactional mode. Cognition sees an opportunity to layer in the conversational fluency and proactive engagement that Poke users have come to expect.

Scott Wu, co-founder of Cognition, described the acquisition as a chance to build an agent that people genuinely enjoy working with, one that is proactive, contextually aware, and capable of sustaining a conversational thread. Wu and fellow co-founder Walden Yan had previously invested in The Interaction Company as angel backers, signaling early interest in the team's approach.

For Cognition, the strategic value extends beyond tone. Poke's architecture is built to handle multiple concurrent conversations and maintain context across sessions. That capability could help Devin manage parallel tasks and remember prior work, addressing a current limitation in which Devin can only process one pull request at a time. Von Hagen suggested that Poke could eventually orchestrate multiple Devin sessions, functioning as a persistent coordinator that tracks ongoing projects and routes requests intelligently.

The Infrastructure Exchange

The deal is structured as a two-way capability transfer. Poke gains access to Cognition's models and infrastructure, which should improve speed and reliability. Cognition, meanwhile, absorbs Poke's interaction design and conversational engine. Starting next year, the Poke team plans to experiment with Cognition's newest software engineering model, SWE-1.7, to handle certain tasks more efficiently.

Nothing will change for Poke users through the end of the year. The assistant will continue operating on Apple's Messages for Business platform and across its existing messaging channels. But in 2027, the two products may begin to converge. Von Hagen noted that fully combining Poke and Devin remains a possibility, though the team is still exploring how best to integrate the two experiences without sacrificing what makes each product useful.

One scenario involves Poke becoming more technically capable by leveraging Devin's coding prowess, while Devin adopts Poke's conversational style and session management. Another involves Poke functioning as a layer that sits above Devin, orchestrating workflows and maintaining continuity across interactions. The exact path forward will depend on how users respond to early experiments and what trade-offs emerge in practice.

A Broader Shift in AI Design

At DailyTechWire, we've tracked a pattern emerging across the region and beyond: startups are beginning to compete not just on model performance but on interaction design. In Seoul, Singapore, and Bengaluru, teams are experimenting with agents that adapt tone based on user preference, context, and even time of day. The hypothesis is that users will tolerate occasional technical limitations if the experience feels human and responsive.

This shift has implications for both consumer and enterprise AI. In consumer contexts, personality can drive retention and word-of-mouth growth. In enterprise settings, it can reduce friction in workflows that require sustained collaboration between humans and machines. Cognition's acquisition of Poke suggests that even in highly technical domains like software engineering, the quality of interaction can influence adoption and long-term engagement.

The challenge, as von Hagen acknowledged, is balancing personality with reliability. Poke's conversational style attracted users, but the cost of running the platform limited its ability to scale profitably. Cognition's infrastructure may solve that problem, but integrating two distinct products without diluting what makes each one valuable will require careful execution.

What Comes Next

The next twelve months will serve as a testing ground. Poke's team will experiment with how Cognition's models can improve task execution while preserving the assistant's conversational fluency. Devin's team will explore how to incorporate Poke's interaction model without slowing down core workflows. Both products will remain operational, giving users time to adjust and providing the combined team with real-world feedback.

If the integration succeeds, Cognition will have a differentiated product that competes on both technical capability and user experience. If it stumbles, the company risks alienating existing users of both platforms. The stakes are high, but the bet reflects a broader industry trend: as AI models become more commoditized, the interface and interaction layer may become the primary source of competitive advantage.

For now, Poke users can continue texting their AI assistant as usual. Developers using Devin will see changes roll out gradually, starting next year. And the rest of the industry will be watching to see whether personality can truly move the needle in enterprise AI adoption.