Cognition just paid a sum in the low nine figures for a startup whose main product chats with users like an old buddy. The move goes beyond adding another tool to its kit. It marks a calculated bet that the way artificial intelligence talks matters as much as what it can do.
The acquisition of The Interaction Company of California, creator of the text-based AI assistant Poke, closed this week. TechCrunch first detailed the transaction. Terms remained private but the valuation landed in that low nine-figure range. Cognition co-founder Scott Wu had skin in the game already. He and his counterpart Walden Yan backed the company as angel investors.
Poke launched in March. Users fire off requests through iMessage, SMS, Telegram or WhatsApp in select markets. They ask it to handle travel plans, health questions, financial chores, scheduling or study help. Productivity wins out as the top category. Managing emails, setting reminders and clearing to-do lists keep people coming back.
Numbers tell part of the story. In the past three months alone, users traded more than 100 million messages with the service. Hundreds of thousands have tried it. Yet the operation burned cash. Running those conversations at scale proved costly. Profit stayed out of reach. Marvin von Hagen, co-founder of the Interaction Company, laid out the economics in his interview with TechCrunch.
But users stuck around. They liked the tone. Poke doesn’t bark orders or spit back stiff replies. It talks back with slang, cracks jokes and remembers context like a friend would. That warmth set it apart from the average chatbot. And Cognition noticed.
Its own creation, Devin, ranks among the most advanced coding agents available. The system writes code, opens pull requests and handles software engineering tasks with impressive skill. One limitation persists, however. Devin completes one pull request at a time. It lacks persistence across sessions. It doesn’t feel like a teammate who sticks around.
Enter Poke. The plan calls for grafting its conversational model onto Devin. The coding agent could gain humor, familiarity and a sense of ongoing partnership. “You probably prefer it if you have co-workers that have personality, rather than if you have co-workers that are just robots,” von Hagen told TechCrunch. “When you have co-workers that are also software engineers, they can also make a joke. And you’ll find it funny and enjoyable.”
Scott Wu echoed the sentiment in a company blog post. “The Interaction team has built an agent that people love: it’s proactive, it knows you, and it’s fun to talk to. That’s exactly how working with Devin should feel, and now we get to build it together.”
The deal arrives at a moment when many AI builders chase bigger models and faster inference. Raw capability still counts. Yet a growing crowd argues that interaction quality determines whether people reach for the tool daily or abandon it after a few tries. Personality, in this view, drives retention. It turns one-off queries into sustained collaboration.
Recent coverage backs the trend. ExplainX noted that Cognition completed multiple acquisitions in quick succession, including the TierZero team focused on post-deployment automation. The pattern suggests an aggressive push to broaden beyond pure coding into persistent, user-facing agents. KuCoin’s flash report added that the purchase marks Cognition’s second deal in three days, highlighting rapid consolidation across AI and crypto circles.
Poke earned another distinction in June. Apple approved it as the first AI agent on the Messages for Business platform. That integration gave businesses a standardized channel inside the Messages app. Users will see no immediate changes through the rest of this year. Poke stays on Apple’s service. Experiments begin next year.
The Interaction team plans to tap Cognition’s latest software engineering model, SWE-1.7, for selected tasks. Speed and reliability should improve. In return, Devin stands to absorb Poke’s proactive style and memory across sessions. Von Hagen sketched a future where Poke orchestrates multiple Devin instances. “I think in the long term Poke can be more reliable with orchestrating all these coding tasks…and, at the same time, Devin could become more like Poke,” he said. “Like right now, Devin can only do one PR — pull request — at a time. I think there’s a lot of value in having Poke orchestrate different Devin sessions.” He summed it up neatly. “It would be good to have a persistent co-worker.”
Observers on X reacted with a mix of excitement and skepticism. Some users joked about promised shares or wondered whether jokes during code reviews would annoy more than amuse. Others saw validation. One post captured the core idea: Cognition bought Poke because it figured out that AI personality beats pure capability in certain contexts. The thread noted the 100 million messages and the cash burn, calling the acquisition a practical fix for both vibe and viability.
This isn’t the first time personality has surfaced in AI discussions. Earlier threads on proactive agents stressed the need for systems that sift data, pick priorities and act without constant prompting. Poke already shows signs of that behavior. It doesn’t wait for explicit commands on every reminder or follow-up. It knows the user. It stays engaged.
Critics still ask tough questions. Will developers want their coding partner to banter while debugging at 2 a.m.? Does adding humor risk diluting focus on accuracy? Cognition seems prepared to test those boundaries. The low nine-figure price tag suggests it views the upside as substantial.
Integration details remain under wraps. Full merger of the products hasn’t been ruled out. For now the companies will run experiments. Poke gains better infrastructure. Devin gains a warmer voice. The combination could produce an AI colleague that not only ships code but also remembers yesterday’s decisions, flags related tasks and occasionally lightens the mood with a well-timed remark.
Industry watchers have begun to track similar moves. Reports from PANews and Phemex emphasized the alignment with Cognition’s goal of building proactive, personalized agents. The acquisition gives the startup direct access to hundreds of thousands of consumer users while strengthening its enterprise coding offerings.
Financial pressure on standalone consumer AI apps like Poke has mounted. High inference costs eat margins even at scale. Teaming with a well-funded coding specialist offers a lifeline. It also hands Cognition a proven interaction layer it couldn’t build from scratch overnight.
Longer term, the deal raises bigger issues about what defines leadership in AI. Model size still dominates headlines. Training runs grow more expensive by the month. Yet if users increasingly choose tools that feel human, then companies ignoring tone, memory and personality may lose ground regardless of benchmark scores.
Cognition clearly believes the two elements must converge. A brilliant coder that feels cold and forgetful will lose to one that remembers your preferences and delivers feedback with a touch of wit. The Poke acquisition accelerates that convergence.
Users can keep texting Poke as before. Developers can keep pushing Devin on complex repos. Behind the scenes, the two systems will start to learn from each other. The result might look less like software and more like that reliable colleague who gets the joke and ships the fix.
Whether the bet pays off will unfold over the next year of experiments. One thing looks clear already. In the race to build indispensable AI, conversation skills have moved from nice-to-have to must-have. Cognition just placed a sizable wager on that shift.
WebProNews is an iEntry Publication