AI Companies Are Buying Personality, Not Just Models

Two odd deals landed the same week, and they say more about your roadmap than any model release did. An image-generation lab bought an astrology app. A coding startup paid nine figures for a chatbot you text like a friend. Neither buyer needed more compute. They needed a voice. Let me catch you up on why that matters for what your team ships next.

The models stopped being the point

Cognition paid a price valuing Poke in the "low nine figures," and they were not buying a better model. They already build models. They bought how Poke talks. It chats like a friend, cracks jokes, and uses slang, and that is the thing Cognition wanted to graft onto Devin, their coding agent.

Cognition co-founder Marvin von Hagen put it plainly: "You probably prefer it if you have co-workers that have personality, rather than if you have co-workers that are just robots." The goal is to make Devin feel like a colleague, not a tool. When the underlying models are close to a tie, the way your product speaks is what people remember.

Design got a seat at the acquisition table

Midjourney's Co-Star buy reads strange until you see the org chart move. Co-Star founder Banu Guler becomes Midjourney's Chief Design Officer, and her team is set to build Midjourney's first real apps, starting with one for image generation. Midjourney has no standalone app today. It lives on the web and inside Discord.

So a lab with roughly 4.3 million reasons to stay technical went and hired a consumer design team with a track record of getting people to open an app every day. The bet is that taste and feel travel better than features do. You can copy a model. Copying why 4.3 million people check an app for their horoscope is harder.

Even Meta stopped pretending voice is decoration

Meta spent last year saying it would skip productivity. CPO Chris Cox told employees the plan was "entertainment" and "connection with friends," not "focusing obsessively on productivity" like OpenAI and Google. Now Meta AI plans your calendar, browses Marketplace, and writes daily briefings.

The reversal tells you the productivity race got crowded fast, and features alone no longer separate anyone. When four chatbots can all read your calendar, the winner is the one people actually like talking to. That is a design problem, not a research one.

Personality without reliability just annoys people

Here is the counterweight. Voice buys you nothing if the thing gets the job wrong. Vertu's $6,880 Alphafold ships an eager AI agent called Hermes that acts fast and confident. In testing against a Samsung Galaxy Z Fold 7, Hermes set a reminder for 9:08 p.m. when the request came in at 2:32 a.m., and booked a trip for the wrong dates.

Google's Gemini was less charming and more useful. It asked a follow-up question, confirmed the airport, then got the details right. Confidence with wrong answers reads worse than a plain tool that works. Personality is a multiplier on trust you have already earned, not a substitute for it.

The deep cut

Notice what these buyers did not do. They did not add "personality" as a feature ticket. They put the people who own voice near the top. Guler is a C-level officer, not a contractor. Poke's interaction team joined to reshape how Devin behaves, not to skin it.

So the move for you is structural. If your brand voice and character live in a copy doc a PM opens twice a year, you are treating as decoration the thing these companies just paid nine figures for. Put voice and character work in the same review as your core flows, owned by someone senior, judged by whether people come back. And pair it with a reliability bar, because a chatty agent that books the wrong flight loses faster than a boring one that works.

Three questions for your team

1. Who owns our product's voice and personality right now, and are they senior enough to say no in a design review?

2. If a rival matched our model quality tomorrow, what about how our product talks would make someone stay?

3. Where does our assistant act confident and get it wrong, and are we shipping charm before we have earned the trust to carry it?