Midjourney bought the astrology app Co-Star: A Practical Take for Solo Builders

An AI image lab just bought an astrology app, and I think it’s one of the more quietly useful signals of the year. The Verge reported that Midjourney acquired Co-Star — the app whose daily horoscopes people actually read — with founder Banu Guler staying in charge of Co-Star while also becoming Midjourney’s chief design officer. Her team’s job: build Midjourney’s first real apps, including one dedicated to image generation, as the company finally outgrows its Discord-and-browser home. Terms weren’t disclosed; the deal reportedly closed in spring.

If you read my piece on the Asana–StackAI acquisition, you’ll recognize the shape of the move — an established platform buying a smaller team for what it does better than them. But this one points the opposite direction, and that inversion is the whole story. Asana bought capability. Midjourney, which already owns world-class capability, bought taste — a consumer product with a voice, a personality, and a reason people open it every single morning. (And if you’ve read my take on why all AI output sounds the same, you already know taste is the exact thing models are worst at generating and audiences reward hardest.)

Reading the deal like a builder

Strip the astrology away and look at what changed hands. Co-Star’s product does three things models don’t: it has a voice you’d recognize without the logo, it manufactures a daily return (the push notification people don’t hate), and it sells feeling — the horoscope lands because it’s personal, not because it’s accurate. Those are design muscles, and Midjourney just hired the whole gym.

Now line that up with the openai family-of-devices play I covered recently: the big labs are all converging on the same conclusion — the model layer is commoditizing, and the winners of the next phase own the surface where people meet the model. OpenAI is betting on hardware. Midjourney is betting on consumer app craft. Same diagnosis, different prescription. When two labs that don’t agree on anything else agree on this, update your priors: the interface layer is where the next fight happens.

There’s also a practical subtext for anyone who generates images for a living-ish. Midjourney has lived inside Discord and a web dashboard since launch — brilliant for power users, invisible to normal humans. A dedicated app designed by someone who made checking your horoscope a daily ritual tells you exactly who that app is for: the mass market. Expect onboarding that doesn’t require reading docs, templates before raw prompts, and a push-notification loop around your images. OpenAI made the same consumer-app bet with its own family of devices — different hardware, identical conclusion.

One more layer underneath: AI-generated imagery is becoming consumer furniture, not just a builder’s tool. Between AI influencers with real followings and product photos nobody can tell were generated, the audience that meets these images has stopped asking where they came from — they just judge whether the experience felt personal. That’s the judge Midjourney just hired a designer to win over.

What I’d actually do about it

1. Treat design as your moat, not your coat of paint. You cannot out-model Midjourney and you cannot out-price it. But nobody at Midjourney knows your niche the way you do. Co-Star won in astrology — a category with a thousand free options — because its voice made it feel like it was written for one person. Whatever you’re building for your audience, the voice and the packaging are now the product, not the garnish.

2. Steal the habit loop, whatever you build. Look at what Guler’s team engineered: a reason to come back tomorrow. If your thing is a newsletter, a template pack, a client dashboard — what’s the daily or weekly return? Most solo products are built as destinations and then wonder why nobody visits twice.

3. Watch for the image-app surface change, then move fast. When the Midjourney app lands, the workflow for non-designers shifts: less prompt-craft, more product. That’s when you refresh your asset pipelines — I compared the current image tools in which AI image tool actually works, and the ranking reshuffles every time a new surface ships. First-mover content about the new workflow (the tutorial everyone else is too slow to make) is the easiest authority win available.

4. Sell the feeling, price the outcome. Co-Star monetized an emotional ritual, not information you can Google. The lesson generalizes: solo builders compete on how the work feels — the confidence, the taste, the done-for-you calm — far more than on raw capability, which the labs keep giving away closer to free.

The bottom line

Midjourney buying an astrology app sounds like a joke until you see it plainly: the era where a better model automatically wins is ending, and the era where the best experience around the model wins is starting. That shift is the best news a solo builder has gotten all year, because experience is something you can build from a kitchen table without a GPU budget. When the capability giants start buying taste, they’re telling you what they lack — and what you can own.

New here? The start here guide is the beginner path, and the AI Tool Advisor helps you pick the right tools for whatever you build next.