Cognition, the company that makes an AI coding assistant called Devin, has bought The Interaction Company of California, which makes a popular chatbot assistant called Poke. The price was described as "low nine figures," meaning somewhere between roughly $100 million and $300 million. Poke co-founder Marvin von Hagen confirmed the price in an interview with TechCrunch. Cognition co-founder Scott Wu announced the acquisition on the company's blog on July 23, 2026, in a post titled "Welcoming The Interaction Company" (TechCrunch).

Poke launched in March 2026 and lets people chat with an AI assistant through texting apps like iMessage, SMS, Telegram, and, in some markets, WhatsApp. In the three months before the deal was announced, Poke users sent more than 100 million messages on the platform. In June 2026, Poke became the first AI assistant approved to run on Apple's Messages for Business platform. No changes are planned for Poke through the end of 2026, and it will stay on Apple's platform. Starting in 2027, the combined team will experiment with using Cognition's own AI model, called SWE-1.7, for some of Poke's tasks. Fully merging the two products has not been ruled out (TechCrunch).

The strategy comes down to personality. Poke's conversational style and personality will be added to Devin, Cognition's coding assistant. In return, Poke will get access to Cognition's technology to become faster and more reliable. The deal reflects a belief, reported by TechCrunch, that how an AI assistant talks and interacts is becoming just as important to its success as how smart the AI model behind it is (TechCrunch).

The broader context here is worth considering. For most of the past two years, AI companies competed mainly on technical measurements: how powerful their models were, how much text they could process at once, and how cheaply they could run. Those things still matter, but the leading AI models have grown similar enough in raw capability that those numbers are no longer enough to stand out. The new differentiator is how an AI assistant communicates, how it keeps track of a conversation across different apps, and how it maintains a consistent personality that people actually want to talk to over time. Poke's 100 million messages in a single quarter suggest there is real demand for a chatbot that feels like it has a personality rather than sounding like a generic robot.

The financial side of the deal also tells a story. Poke was expensive to run and was struggling to make a profit, according to von Hagen. Handling tens of millions of messages across multiple chat apps costs a lot in computing power, and it is still unclear how companies in this space should make money. Joining Cognition gives Poke access to better technology that could lower its costs, while Cognition gets a chatbot system it can build into Devin. Whether the savings actually materialize depends on how efficiently Cognition's systems can handle Poke's traffic and how quickly the team can put Cognition's model to work on Poke's tasks next year.

The two companies already had a relationship. Cognition co-founders Scott Wu and Walden Yan had previously invested their own money in The Interaction Company, so the acquisition is an extension of an existing financial relationship rather than a deal between strangers. That prior familiarity likely made the negotiations smoother and means the two teams already understood each other's approaches to building products (TechCrunch; Cognition blog).

For Cognition, the bet is that a coding assistant needs more than just the ability to write good code. If developers are going to work with an AI assistant over long sessions, the quality of that back-and-forth communication becomes part of what makes the product worth using. Poke's experience designing conversations across iMessage, Telegram, SMS, and WhatsApp applies directly to that problem. The fact that Cognition is willing to spend hundreds of millions to buy that capability rather than build it themselves suggests the interaction layer is harder to create than it looks, or that getting it done quickly matters more than saving money, or both.

The decision to keep Poke running on its own through the end of 2026, with integration starting in 2027, is a careful approach. It keeps Poke's spot on Apple's Messages for Business platform, which required its own approval process, and it avoids disrupting a service that people are actively using while the teams figure out how the two products should work together. The option to fully merge remains open if the early experiments go well.

What this deal makes possible, if the integration works, is a coding AI assistant that can not only handle software engineering tasks on its own but also talks with developers in a way that feels natural across the apps they already use every day. That is a different kind of promise from simply generating better code, and it positions Cognition to compete on the full experience of working with an AI rather than on technical numbers alone.