Cezar Grigore.

Thoughts

Long thoughts that needed a few thousand words, and notes that arrived fully formed. Newest first.

2026

  1. Note

    I really think people aren't ready yet for the AI that does it for you. And I don't think the AI is ready to do it for you either.

    A lot of companies are going all in on this right now. Meta with Muse, Instinct, everyone. The pitch is the same: I'll book you the right hotel, I'll find you the right experience, I'll buy you the right clothes. A proper assistant that just does it. Yeah, sure, there's a first mover advantage if you manage to capture that. And sure, some of it works really well even today. Finding a plane ticket is simple stuff. But when it fails, the failure mode is so annoying. It's my money, it's the wrong ticket, it's an expensive mistake. I'm just not sure most people want that in their lives yet.

    I think they want something else. There's a reason the smartphone got so big and is still so big. People are visual. They want to see things before they commit. They want to explore. I don't think "I'll do it for you" is what people are ready for today. It'll work great for a small percentage of people. I'm just not sure that's where you start.

    And your life isn't just yours. It's shared. With your partner, with your friends, with your family. The flat, the car, the mortgage, the pet. Finding a restaurant for Saturday, finding a cool place to go in October. That's what your daily life actually is. It's not someone booking things for you in the background. It's a lot of small decisions, and most of them you make together with other people.

    I think you start by getting people to save the stuff they care about. It's the easiest thing to ask of someone, you understand them better, and you build memory. The technology needs that to get to real understanding, and we're not there. Think about it. Meta knows more about you than anyone. You open Instagram and sure, you get some good recommendations. But you still scroll through ten reels before you find one you actually want to double down on.

    So yeah. It's tricky.

  2. Note

    Someone sent us feedback last week and I couldn't work out what they meant. I pasted it into ChatGPT, went back and forth a few times, and there it was: a good idea, buried in a sentence I'd almost skipped past.

    That's backwards. I had a conversation about their feedback with a machine, when the person who wrote it was right there. The form just had no way to ask them anything.

    A feedback form should be able to ask a follow-up. It should know what we're already planning and what other people have asked for, and be able to say: do you mean this, or this? Would that fix it? Almost nobody can write a good feature request. Almost everyone can answer three questions about the thing they're trying to describe.

    Feedback forms have been suggestion boxes for thirty years because listening properly didn't scale. It does now.

  4. Note

    We spent over a year getting Muse to something a normal person would actually use. Not an entrepreneur, not someone who lives in their email. A teacher, a plumber, a doctor. Three or four versions to get there, most of it squeezed around Social Cat and Clip, and Oscar arriving in the middle of it. Finding the name took months on its own.

    This week Meta shipped an assistant called Muse. Ours is done. Not legally, maybe. Practically: nobody is going to find us standing behind that.

    A year of product survives a week like this. A name doesn't.

  5. Note

    I increasingly think the best consumer products aren't those that give you more things to do. They're the ones that quietly remove things you used to have to think about.

  6. Note

    Meta optimizing for installs isn't finding customers cheaply. It's finding installers cheaply. Those are very different things.