Cognition Acquires the Company Behind Poke

SAN FRANCISCO, Calif. — In the rapidly evolving landscape of artificial intelligence, the transition from reactive chatbots to proactive agents represents the next significant frontier for Silicon Valley’s power players. Cognition, the high-profile startup that recently upended the software engineering sector with its autonomous coder Devin, has signaled its intent to lead this shift through a strategic expansion. The company has officially acquired The Interaction Company of California, the Palo Alto-based startup responsible for Poke, a personal AI agent that operates primarily through text messaging.

While the companies did not formally disclose the purchase price, the deal underscores the massive capital currently flowing into the agentic AI space. Industry observers are treating the acquisition as a major consolidation of talent and technology, particularly after TechCrunch reported that the transaction valued Interaction in the low nine figures. This valuation highlights the premium currently placed on startups that have successfully solved for user engagement and long-term memory in AI interactions—capabilities that remain elusive for many of the world’s largest large language model providers.

The move marks a notable bridge between two disparate sectors of the AI market. Cognition is best known for Devin, an autonomous software-engineering agent that represents the pinnacle of technical task execution. Devin is not merely a pair programmer; it is a system capable of planning and executing complex coding work in cloud environments with minimal human intervention. Until now, Cognition’s focus has been largely technical and workflow-oriented. By bringing The Interaction Company into the fold, Cognition is absorbing a company that has mastered the consumer-facing, conversational side of the equation.

Poke represents a departure from the conventional, query-response loop that defines tools like ChatGPT or Claude. It is designed to act more proactively than a conventional chatbot, moving away from the "wait for input" model that has characterized the first wave of generative AI. Poke can initiate messages, follow up on earlier conversations, and help users manage tasks through a familiar messaging interface. This proactive nature addresses one of the primary friction points in digital productivity: the cognitive load of remembering to check in on a project or follow up on a task.

The scale of Poke’s adoption prior to the deal suggests a significant appetite for this style of interaction. Cognition said people exchanged more than 100 million messages with Poke during the three months before the acquisition. For a startup out of Palo Alto, such volume provides a massive dataset regarding how humans interact with proactive software. It demonstrates that users are not only comfortable with an AI reaching out to them via SMS but are actively engaging with those prompts at a frequency that rivals major social media platforms.

Cognition’s leadership viewed the acquisition as a natural evolution of their core philosophy. The buyer said both products are built around the idea of always-on agents that remember context and take action. In the world of enterprise software, context is everything. An engineering agent that understands the history of a codebase is useful, but an agent that can intuitively reach out to a developer to clarify a requirement or report a completed task via a text message brings a level of seamlessness that traditional development environments lack.

To that end, the integration of these two teams has clear immediate objectives. Interaction's team will now work with Cognition on making Devin feel more responsive and easier to communicate with. For Cognition, the challenge with Devin has always been the "black box" nature of autonomous work. While Devin can navigate cloud environments and write code independently, the human-to-agent interface remains a critical bottleneck. By applying the conversational design principles behind Poke to the technical powerhouse of Devin, Cognition aims to make its software engineer agent feel less like a remote script and more like a collaborative teammate.

However, the combination joins two very different kinds of agent experience: a technical tool used for software development and a consumer-facing assistant built around personality and conversation. Devin is a productivity engine designed for accuracy and execution, whereas Poke was built to inhabit the personal space of a user's text message inbox, relying heavily on a sense of personality to drive engagement. This cultural and functional gap presents both an opportunity and a significant strategic hurdle for Cognition as it integrates its new acquisition.

Industry analysts suggest that this dual-track approach could help Cognition improve engagement across its user base, but it also introduces product questions about whether Poke will remain independent or be folded into Devin. If Poke continues as a standalone product, Cognition enters the crowded consumer assistant market, competing with native smartphone integrations and specialized lifestyle bots. If it is folded entirely into Devin, Cognition risks diluting the "proactive" magic of Poke by tethering it strictly to the rigors of software engineering.

The lack of clarity regarding the roadmap is currently a point of speculation among the Pal Alto startup community. Cognition did not provide an integration schedule or explain how Poke's existing service will change for its current millions of users. For those who have come to rely on Poke for task management and proactive check-ins, the acquisition creates a period of uncertainty. For the engineering teams at Cognition, the merger represents a high-stakes experiment in user experience design.

The ultimate test for this acquisition will be whether Interaction's expertise in proactive messaging improves a professional agent without making it distracting or unpredictable. In a professional environment, an agent that "initiates" too often can quickly become digital noise, echoing the worst traits of early workplace notification systems. If Devin begins texting developers with the same frequency that a personal assistant might text a consumer, the line between helpfulness and annoyance becomes thin.

Furthermore, the shift toward proactive AI brings a new set of ethical and operational challenges that Cognition will have to navigate carefully. Trust, consent and clear controls will be especially important when software is designed to contact users before they ask for help. When an agent like Poke or a revamped Devin takes the initiative to start a conversation, it is essentially asserting itself into the user's attention. Maintaining the user's trust while granting an AI the permission to be "always-on" requires a delicate balance of security and transparency.

As the low-nine-figure deal settles, the Silicon Valley tech scene is watching closely to see if Cognition can successfully merge the DNA of a high-end engineering tool with a high-engagement social interface. If successful, the new, more communicative Devin could set a new standard for how humans interact with autonomous labor. If the integration falters, it may serve as a cautionary tale about the difficulties of mixing technical utility with consumer-oriented personality. For now, Cognition stands at the forefront of the agentic revolution, armed with the millions of messages and the proactive specialized technology that The Interaction Company built in the heart of Palo Alto.