Opine Lets You Ask MrBeast for Life Advice, and the Answers Are Surprisingly Good

The Macro: AI Has an Opinion Problem

Ask ChatGPT whether you should quit your job. You will get a balanced, hedge-filled response that lists pros and cons and concludes with something like “ultimately, it depends on your personal circumstances.” Ask it whether pineapple belongs on pizza. Same structure. Same hedging. Same refusal to commit.

This is by design. OpenAI, Anthropic, and the other major model providers have spent enormous effort making their chatbots neutral. The goal is to avoid controversy, reduce liability, and appeal to the widest possible audience. It works for general-purpose use cases. It is terrible for anyone who actually wants advice.

Think about how humans seek advice in real life. You do not go to someone who lists every possible perspective with equal weight. You go to a friend with strong opinions, a mentor who has been through it, a podcast host whose worldview you respect. You want someone who will tell you what they think, not what every reasonable person could think.

The AI persona market has been circling this idea for a while. Character.AI lets you talk to fictional versions of real people, but the conversations are entertainment, not utility. Pi from Inflection was supposed to be emotionally intelligent, but it still hedges. The entire crop of “AI companion” apps is designed to be agreeable, which is the opposite of opinionated.

Nobody has really built for the use case of “I want genuine advice from someone whose judgment I trust, delivered at scale by AI.” That is a specific product with a specific audience, and it is different from chatbots, companion apps, and character role-play.

The Micro: Imperial College Meets EPFL

Opine was founded by Manuj Mishra and Diogo Valdivieso, out of London. Manuj ranked number one at Imperial College London in Mathematics and Computer Science. Diogo previously built Battle Arena, which hit one million users, and holds a CS degree from EPFL. They are part of Y Combinator’s Spring 2025 batch with Tom Blomfield as their partner.

The product gives you access to over 100 AI personas modeled after real public figures. YouTube creators like MrBeast. Podcast hosts like Lex Fridman and Joe Rogan. The personas are built from their public content, so when you ask “Lex Fridman” about career advice, the answer draws from things Fridman has actually said on his show. Responses come with source citations, which is a nice touch and a meaningful differentiator from Character.AI’s “vibes-based” character simulation.

The distinction matters. Character.AI personas are improvising. They take a character description and riff. Opine personas are grounded in actual content. The AI is not guessing what Joe Rogan would say about cold plunges. It is synthesizing what he has actually said about cold plunges across hundreds of podcast episodes.

The interface is a chat-based discovery platform. You can browse available personas, pick one, and ask questions about career, relationships, health, or whatever else you are thinking about. The core promise is right there in the name: these AIs have opinions. They will tell you what they think instead of listing every possible angle.

Two-person team. London-based. Consumer AI. That is a tough combination for fundraising and distribution, but the product concept is differentiated enough to generate organic pull if the persona quality is high.

The Verdict

I think Opine is onto something real. The AI opinion gap is a genuine problem, and the approach of grounding personas in actual content rather than character descriptions is smart. It turns a novelty product into a utility product. The difference between “talk to a fake Joe Rogan” and “get advice based on everything Joe Rogan has actually said” is the difference between a toy and a tool.

The risks are clear. Licensing is the first one. Using real people’s likenesses and content to build AI personas raises legal questions that have not been settled yet. MrBeast might think it is cool. His lawyers might not. Opine will need to figure out whether they are building partnerships with creators or building on top of public content and hoping nobody objects.

The second risk is depth. A hundred personas sounds like a lot until you start using them and realize that most people will gravitate toward five or ten favorites. If those five are not good enough to keep you coming back, the breadth does not matter. Better to have ten incredible personas than a hundred mediocre ones.

Thirty days, I want to know which personas get the most usage and whether people are asking real questions or just testing the novelty. Sixty days, whether retention holds after the first session or drops off like every other AI chat product. Ninety days, the big question: can Opine sign creator partnerships that give them exclusive access to content and personas, or will Character.AI and the other incumbents copy the “sourced opinions” model and compete on distribution? The product is good. The moat is still under construction.