Cognition acquiring Poke is a sign that the next generation of AI will not live only in developer consoles and browser chat windows. It will sit inside the messaging threads people already use every day and quietly turn from reactive answer engines into persistent collaborators. By buying The Interaction Company of California, the team behind Poke, Cognition is betting that personality, presence, and context will matter as much as raw model performance in the race to build useful AI coworkers.
AI is moving from browser tabs into our message threads as persistent, context-aware coworkers
From autonomous coding to everyday conversations
Cognition is best known for Devin, a software engineering agent that captured attention earlier in the year by tackling complex coding tasks with more autonomy than traditional code assistants. The company has positioned Devin as an AI software engineer rather than a simple autocompletion tool, signaling an ambition to move beyond prompt response into multi-step work on behalf of developers.
Poke has grown up in a very different context. It is a text-based personal assistant that users interact with the same way they would text a friend, handling planning, calendar management, health tracking, smart home control, and lightweight creative tasks from inside ordinary messaging threads. The agent operates through Apple Messages, SMS, WhatsApp, and Telegram, and has been approved as the first third-party AI agent on Apple Messages for Business.
In just three months, Poke has processed more than 100 million messages from hundreds of thousands of users, and it is currently the only AI agent that can text natively through Apple Messages. That volume and distribution give Cognition access to a proven interaction model and infrastructure for running persistent message native agents at scale.
The acquisition reportedly values Poke in the low nine figures, a striking number for a product that only launched publicly in early 2026 and still operates primarily as a consumer assistant. While Cognition has not disclosed official purchase terms, the reported valuation highlights how much strategic weight investors and acquirers now place on AI interaction design, not just model weights and benchmarks.
What exactly Cognition is buying
Cognition is not simply taking an investment stake. It is acquiring The Interaction Company and bringing the Poke founders directly into its AI coding business. Public announcements from Cognition emphasize that Poke users will retain access to the product and that Cognition’s models and infrastructure will be used to make Poke faster and more reliable over time.
In practice, Cognition is buying three things.
First, it is acquiring a mature interaction model. Poke is built to initiate conversations, send reminders, and follow up on tasks without waiting for the user to open an app or craft a prompt. That proactive behavior is exactly what most developer agents lack today.
Second, it is acquiring an inbox native architecture. Poke lives directly inside Apple Messages and SMS, and mirrors this behavior where available in WhatsApp and Telegram. This placement gives the agent continuous access to the context of conversations and tasks that people already manage through their phones, from family logistics to work coordination.
Third, Cognition is acquiring a real-world stress test. Running more than 100 million conversations in three months with hundreds of thousands of users is not only a growth story. It is proof that the underlying infrastructure can support constant back-and-forth interactions across carriers and platforms.
When Cognition’s cofounders describe a shared vision of always-on cloud agents, they are pointing to this combination of proactive behavior, inbox presence, and production seasoning. It is the opposite of the traditional chatbot that waits passively for a question in a browser tab.
Inbox native agents and why they matter
The most important shift in this deal is where the agent lives. Poke is also the only AI agent currently approved for native texting on Apple Messages, underscoring how tightly the product is woven into mainstream messaging infrastructure. Poke does not ask people to download a new app, remember a new interface, or change their habits. It appears as a contact in Apple Messages or as a number in SMS, then extends similar behavior into WhatsApp and Telegram where supported.
Because it resides in the same threads that people already use for personal and work communication, Poke can monitor relevant context, send proactive reminders, and follow up on tasks in place. You might reschedule a meeting, log a workout, request a grocery list, or review travel plans without ever leaving your main messaging inbox.
From a product adoption perspective, this is powerful. Traditional productivity apps and chatbots require users to take deliberate action to engage, whether opening a browser tab or switching to a dedicated interface. Message native agents reduce friction by showing up where attention already is. That structure is also well aligned with how human assistants work. They operate in the background, stay aware of ongoing threads, and gently nudge when something needs attention.
For enterprises, inbox native agents suggest a future in which AI coworkers operate across channels rather than remaining locked inside developer environments. Workflows that run through email and chat could be quietly supported by agents that track tasks, deadlines, and code changes without requiring manual check-ins. Poke’s early work on Apple Messages for Business previews the same concept for customer service and business messaging.
Blending Poke’s personality with Devin’s technical depth
Cognition’s stated strategy is to blend Poke’s conversational personality and proactive behavior with Devin’s technical strength in software engineering tasks. That means taking a companion that knows how to feel friendly, persistent, and context-aware, and pairing it with a model that knows how to read repositories, reason about code, and coordinate workstreams.
Poke has been explicitly designed to feel like a friend you text rather than a formal chatbot. It uses casual language and keeps continuity across conversations so that users perceive it as an ongoing presence rather than a series of isolated sessions. Devin, by contrast, is optimized for serious technical work, focusing on code bases, tickets, and engineering tasks.
The acquisition creates an opportunity to fuse these strengths. Cognition has already indicated that Poke will begin to use its newest software engineering model, SWE 1.7, for certain tasks next year, and that the team will experiment with ways both Poke and Devin can improve by sharing capabilities. Fully combining the products is not ruled out, but Cognition has said that Poke will remain on Apple’s platform and that nothing will change immediately for users through the end of the year.
If that roadmap holds, developers could see agents that do more than answer prompts. They could receive proactive messages when tests start failing, when a dependency looks risky, or when a refactor would pay off. Agents could coordinate tasks across chat and issue trackers, nudging teammates and escalating only when needed. Poke’s experience initiating contact and maintaining conversational context is directly applicable to these scenarios.
AI personality as a competitive advantage
Coverage from Yahoo Finance and other outlets has framed this deal as a recognition that AI personality is becoming a competitive edge, not just a design detail. Many assistants now draw from similar underlying model families, whether large language models or specialized coding models. What often differentiates them is how they engage users, how much initiative they show, and whether they feel trustworthy and helpful over time.
Poke’s growth shows that users respond to an assistant that acts like a persistent companion rather than a transactional bot. Processing over 100 million messages in three months is not simply a metric of scale. It suggests that people are willing to let an AI agent live in their daily messaging streams and handle a wide spectrum of tasks from the mundane to the personal.
For Cognition, this is directly relevant to enterprise and developer products. As AI coworkers move closer to day-to-day workflows, they will need to maintain long-running context, handle interruptions, and respect boundaries. AI literacy and behavior will influence whether teams actually trust agents to monitor conversation threads or code changes continuously.
There is also a defensive dimension. Once users are accustomed to a particular agent that lives in their inbox, switching away becomes harder. That stickiness is valuable both for consumer assistants like Poke and for professional agents like Devin. An AI that knows your habits, your communication style, and your work patterns is not easily replaced.
Risks and unanswered questions
Despite the enthusiasm, there are real risks and uncertainties. Cognition has not publicly detailed the full product roadmap for how Poke and Devin will converge, beyond high-level comments about integrating models and infrastructure and keeping Poke available. That leaves open questions about how pricing, branding, and data governance will evolve.
Privacy is a central concern. Inbox native agents, by design, sit close to personal and work communication. Even if they only act on messages explicitly sent to them, the perception of an AI operating inside core channels will require careful communication, clear consent mechanisms, and robust controls. Enterprise customers will want guarantees about how data flows between Poke, Devin, and Cognition’s back-end systems.
There is also the question of reliability. An AI that occasionally makes a mistake when answering a single question in a browser chat is one thing. An AI that proactively reaches into developer workflows or personal schedules carries more responsibility. Failure modes become more serious when agents have the freedom to initiate actions or coordinate work.
Finally, the market landscape is shifting quickly. Other players are experimenting with message native agents, integrated assistants inside productivity suites, and more autonomous coding partners. The fact that Poke appears to be the first iMessage agent to be acquired underscores how early this category still is. Cognition’s move may trigger further consolidation or competition from larger platforms that control messaging ecosystems.
What this means going forward
The acquisition of Poke by Cognition is a meaningful moment in the evolution of AI agents because it connects two important threads. One is the push toward autonomous coding coworkers represented by Devin. The other is the rise of persistent, message native assistants represented by Poke.
Several takeaways stand out.
- Interaction design is becoming strategic. The reported low nine-figure valuation shows that investors now value proven interaction models and personality almost on par with core models.
- Inbox native is likely to spread. Poke’s success inside Apple Messages and SMS suggests that agents embedded in existing communication channels can drive engagement more effectively than standalone apps.
- Enterprise AI will feel more human. As Poke’s conversational style and proactive behavior flow into Devin and related tools, developer and enterprise agents are likely to become more conversational, more persistent, and more context-aware. That shift will make them more useful but also raises new questions about trust and control.
- The line between consumer and professional agents will blur. A single AI presence that helps with personal planning and work tasks from the same messaging threads is now plausible. That will force companies to rethink how they separate and protect different types of data.
Taken together, Cognition’s purchase of Poke signals a broader belief that the future of AI will be defined not only by bigger models but by agents that live alongside us in our daily communication, remember what matters, and take initiative when needed. The next few product cycles will show whether this bet on personality, persistence, and context pays off at scale.
Conclusion
Cognition’s move to acquire Poke is a clear sign that AI agents are shifting from experimental tools to embedded companions living inside the messaging apps people already use every day. It matters right now because it connects a serious software engineering agent in Devin with a consumer friendly messaging agent in Poke, and it does so at the moment when platforms like Apple Messages are opening the door to native AI experiences for hundreds of millions of users.
How we arrived at texting AI like a friend
For more than a decade, AI assistants have mostly lived in separate apps or voice interfaces. Early systems like Siri and Google Assistant sat inside operating systems but were constrained by narrow skills and rigid commands. Enterprise tools focused on automation and coding help while consumer chatbots stayed inside browser tabs or standalone mobile apps.
Messaging agents like Poke represent the next stage in that evolution. Instead of asking people to install yet another app or visit a website, Poke runs directly inside channels such as Apple Messages, SMS, WhatsApp and Telegram, turning the humble text thread into an interface for planning, reminders and light automation. Poke is built by The Interaction Company of California, and over the past few months its interface has processed more than 100 million messages, serving hundreds of thousands of users who interact with it much as they would with a helpful contact.
Apple’s decision to approve Poke as the first third party AI agent on its Messages for Business platform is particularly important. It means a person can now text a standalone AI agent inside Apple Messages in almost the same way they would text a friend or a company, within a service that reaches roughly one billion users globally. Apple bills Poke on a per user basis, a model the startup has suggested compares favorably to Meta’s per message fees for third party AI on WhatsApp, signaling how major platforms are experimenting with economic models for embedded agents.
What Cognition is actually buying
Cognition is best known for Devin, its AI coding agent that aims to behave more like a software engineer than a static autocomplete tool. Devin is designed to take on tasks, manage its own environment and push code, which makes it largely an enterprise and developer focused product.
By acquiring The Interaction Company and Poke, Cognition is buying a proven interaction model for everyday users rather than just a feature set. Reports indicate the deal values Poke in the low nine figures, although Cognition did not disclose official purchase price or terms in its own announcement. TechCrunch previously reported that Poke had raised significant venture funding and was valued around 300 million in an earlier round, which helps explain why the interaction design itself is now seen as a strategic asset.
Poke is a text based assistant that messages users like a friend, initiates conversations, follows up on tasks and manages routines, all within familiar threads. It has already exchanged around 100 million messages in three months, and it operates directly in SMS as well as inside Apple’s Messages for Business where it is the first approved AI agent from an independent developer.
Cognition’s leadership has emphasized that Poke users will continue to access the product without disruption, and that the company plans to use its own models and infrastructure to make Poke faster and more reliable while preserving the current user experience. Poke will stay on Apple’s platform through at least the end of the year, with deeper experimentation and integration expected after that, including use of Cognition’s newest software engineering model SWE 1.7 for some tasks.
From enterprise agent to consumer companion
The strategic logic becomes clearer when you look at how Devin and Poke complement each other. Devin brings sophisticated capabilities in software development and task management, but it is primarily aimed at technical teams and structured workflows. Poke excels at casual conversational engagement and at starting and sustaining threads with non technical users inside everyday messaging apps.
Commentary from investors and analysts has pointed out that Poke’s real asset is its ability to keep messaging relationships alive and useful. One investor noted that Poke exchanged roughly 100 million iMessages with users in its first three months, arguing that the capability to start threads, keep them going and suggest new uses is a powerful revenue driver for usage based products like Devin. When you combine Devin’s enterprise grade skills with Poke’s consumer friendly personality and proactive behavior, you get a blueprint for an agent that can work across both professional and personal contexts without forcing the user to change interfaces.
This is also why several observers describe the acquisition as a bet on AI personality as a competitive advantage. TechCrunch and others have stressed that the deal shows interaction style is now valued alongside model performance, and that assistants people enjoy texting may matter as much as assistants with the strongest benchmarks.
Why messaging matters as the AI control layer
Embedding agents directly into iMessage and SMS matters for three reasons.
First, messaging is already the default control surface for many people. They coordinate projects, social plans and logistics through text threads. An AI that lives in those same threads and quietly tracks tasks, schedules and reminders can slide into existing habits instead of demanding new ones. Poke’s design does exactly this by sending proactive messages, following up when tasks are incomplete and offering help in a conversational tone.
Second, messaging is cross device and resilient. A text based agent works across smartphones, tablets and even basic feature phones in the case of SMS. That makes it more inclusive than desktop only tools or voice assistants tied to a single brand of speaker.
Third, platform level approval such as Apple’s Messages for Business decision gives these agents deep integration. Poke is the first AI agent of its kind to be approved for that interface, which positions it as a template for future agents that want to live inside Messages without feeling like a bolt on bot. Apple’s per user billing model also pushes developers to think about sustained value rather than opportunistic engagement since costs scale with the number of active people rather than raw message volume.
Implications for the technology stack
On the technical side, Cognition plans to migrate Poke onto its own models and infrastructure. The company has said that this will make Poke faster and more reliable while keeping the familiar user experience intact. That suggests a layered architecture where Devin and Poke can share core capabilities such as planning, scheduling and tool use, while presenting different personalities and interfaces depending on whether the user is coding, managing a work project or just organizing daily life.
The integration with Cognition’s latest engineering model SWE 1.7 hints at a future in which Poke can do more than reminders and lightweight tasks. For example, Poke could eventually trigger real development workflows through Devin behind the scenes, or orchestrate more complex automations such as updating systems, generating reports or calling external APIs, all initiated from a simple text conversation.
From an infrastructure perspective, bringing Poke’s high volume messaging traffic onto Cognition’s stack provides real world data on how ordinary users interact with agents day to day. That data, if handled responsibly, can tighten feedback loops around dialog management, error recovery and user satisfaction. It also raises questions about privacy, data retention and transparency, which will be crucial if these agents are to become trusted companions rather than opaque black boxes.
Business and societal impact
For Cognition this acquisition is a bet that everyday interaction is the gateway to sustained usage and revenue. Devin is a usage based product. If agents like Poke can start threads, keep them alive and suggest new uses, they can help drive recurring engagement that translates into more work done and more value captured on Cognition’s platform. The reported low nine figure valuation underscores that part of the market now prices conversational design and personality alongside raw model capability.
For businesses, the move points toward a future where employees and customers increasingly work through agents inside standard messaging apps rather than bespoke dashboards. A manager might text an agent to coordinate tasks across a team, while a customer might text a brand’s AI to track an order or solve a problem. Poke’s approval on Messages for Business shows that platforms are willing to let independent agents sit next to corporate accounts, which could blur lines between consumer assistant and business helper.
For society, the spread of always on cloud agents inside personal messaging threads has both clear opportunities and real risks. On the opportunity side, people who find traditional productivity tools overwhelming could benefit from a companion that nudges them about appointments, habits and goals in simple language. Poke’s early adoption suggests that texting an assistant like a friend can make advanced AI feel less intimidating and more approachable.
On the risk side, the more private and continuous these interactions become, the more critical issues of consent, data protection and agency feel. A proactive agent that initiates conversations and keeps detailed records of routines could be immensely helpful, but it could also become a vector for subtle manipulation or overdependence if incentives are not aligned with user wellbeing. Platform rules, billing structures and transparency reports will matter as much as model architecture in determining whether these companions remain trustworthy.
There is also platform risk. Apple’s per user billing model and its gatekeeping role on Messages for Business give it enormous influence over which agents can scale and on what terms. Developers may find themselves balancing innovation against dependency on one or two dominant messaging platforms, in much the same way that mobile app developers had to adjust to store policies a decade ago.
What to watch next
Over the next year several questions will determine how transformative this acquisition really is.
One is how deeply Devin’s capabilities flow into Poke without overwhelming its casual feel. If Poke can quietly tap into sophisticated planning and coding skills while still feeling like a friendly contact, it could set the pattern for future agents that blend serious work with everyday conversation.
Another is how Apple and other platforms respond. Poke’s status as the first approved AI agent in Apple Messages for Business sets a precedent, and further approvals or new policies will show whether platform owners intend to cultivate a diverse ecosystem of agents or keep tight control over a small set of partners.
Finally, it will be important to see how Cognition handles transparency around data and behavior. Clear documentation, user controls and visible boundaries between enterprise workflows and personal messaging will be essential if AI companions are to become a normalized part of daily life rather than a passing novelty.
The core takeaway is that Cognition’s acquisition of Poke marks a turning point in how AI agents are delivered. Rather than living in isolated tools, they are moving into the center of everyday communication channels and learning to behave more like enduring relationships than single use apps. If the integration succeeds, millions of people could find that the most capable assistant they have is not a new device or service, but a familiar text thread where an almost invisible companion quietly keeps life running in the background.
