Cognition has acquired ‘Poke’, an AI assistant built around casual text-message interaction, in a move that shows how much AI companies are starting to care about product personality, not only raw model performance.
The deal brings ‘Poke’ into the same company behind Devin, Cognition’s AI coding agent. The company says the goal is to make Devin feel less like a rigid tool and more like a useful teammate that can understand context, remember work, and communicate in a more natural way.
That matters because the AI assistant market is becoming crowded. Many products can now answer questions, write code, summarize documents, and handle basic task workflows. The harder question is whether users actually want to keep working with a specific assistant every day.
Tech Block Central has recently covered how new AI models are competing on capability, including Anthropic’s ‘Opus 5’ launch. Cognition’s ‘Poke’ acquisition points to a different part of the same fight: how the assistant feels when people use it.
Why Cognition Bought ‘Poke’
‘Poke’ was developed by The Interaction Company of California and launched earlier this year as an AI assistant users could message like a friend. Instead of forcing people into a traditional app interface, it works through messaging channels and uses a more informal style, including humor and casual phrasing.
That style appears to be the reason Cognition wanted the team. The company already has a coding agent in Devin, but coding work is not only about completing tasks. Developers also need an assistant that can explain what it is doing, ask useful follow-up questions, keep track of ongoing projects, and avoid feeling like a black box.
Cognition co-founder Scott Wu described ‘Poke’ as proactive, personal, and enjoyable to talk to. In practical terms, that suggests Cognition wants Devin to become more persistent and context-aware, rather than just a coding agent users prompt one task at a time.
The acquisition also gives ‘Poke’ access to Cognition’s models and infrastructure. That could make the assistant faster and more reliable, while giving Cognition a consumer-style interaction layer that it can bring into developer workflows.
AI Personality Is Becoming A Product Feature
For years, AI companies have mostly competed on benchmarks, context windows, speed, pricing, and model quality. Those still matter. But as more products become capable enough for everyday work, personality and interaction design are becoming harder to ignore.
A coding assistant that completes a task but communicates poorly can still slow a developer down. A productivity assistant that technically works but feels awkward may be abandoned after the novelty fades. That is why the tone, memory, and rhythm of an AI assistant can become a real advantage.
‘Poke’ is interesting because it treated the assistant as a relationship-shaped product from the beginning. Users interacted with it through familiar messaging apps, and the product leaned into a more human style. That can be risky if the assistant overpromises or becomes distracting, but it can also make AI feel less like software people have to manage.
The challenge for Cognition will be keeping that personality useful inside serious coding work. Developers do not need forced jokes when a production bug is on the line. They need clarity, context, and confidence that the assistant understands the task.
What This Could Mean For Devin
Devin is designed to operate more like an AI software engineer than a simple code-completion tool. If Cognition brings ‘Poke’-style interaction into Devin, it could make the product more conversational across longer projects.
That could include remembering prior tasks, coordinating multiple work sessions, explaining tradeoffs more naturally, and helping users understand what the agent is doing without requiring constant supervision.
One possible direction is that ‘Poke’ could sit above several Devin sessions and help coordinate them. For example, a user might ask the assistant to track several code changes, remember blockers, and keep different tasks organized over time.
Cognition has also introduced its SWE-1.7 model family, and ‘Poke’ may eventually use more of that infrastructure. That would make the acquisition more than a branding move. It could become a bridge between consumer-like AI interaction and technical agent workflows.
Why It Matters For The AI Assistant Market
The acquisition suggests AI companies are moving beyond a narrow focus on who has the strongest model. The next layer of competition may be who can make an assistant that people trust, understand, and keep using.
That is especially important as AI agents take on more open-ended tasks. The more autonomy an assistant has, the more users need it to communicate clearly. Personality is not just decoration in that context. It can affect whether the product feels dependable or frustrating.
There is also a business reason for this shift. Running AI assistants can be expensive, especially when users send a large number of messages. Products need strong retention and clear usefulness to justify those costs. A better interaction model can help keep users engaged, but only if the assistant is genuinely helpful.
For Cognition, ‘Poke’ gives Devin a clearer path toward becoming a more personal work companion. For the wider market, it is another sign that AI agents will be judged on product experience as much as model strength.
The open question is whether users want AI assistants with strong personalities inside professional tools, or whether they mostly want quieter agents that stay out of the way. Which version would you trust more for real work?
Further reading: Cognition’s announcement and Cognition’s SWE-1.7 model update.