Last quarter, we built a feature at Murror that our engineering team jokingly called "the empty room." After a user finishes a journal entry, the AI doesn't immediately respond. It waits. For 30 seconds, the screen shows nothing but the user's own words and a gentle prompt: "Sit with what you just wrote."
No analysis. No reframe. No pattern recognition. Just silence.
High-risk got pushed to December 2027 and the whole timeline read as "delayed." Article 50 didn't move. It applied on 2 August, it catches you if an EU user sees your output, and the ceiling is 15m or 3% of worldwide turnover.
Two and a half weeks ago a rule started applying to most people reading this, and the coverage around it said the opposite.
When we first built Murror's reflection AI, we optimized for insight. Every journal entry got a thoughtful, confident analysis. Pattern recognition, emotional connections, suggestions for growth.
AI apps convert trials 52% better and churn annual subscriptions 30% faster. The flattering metric lands in week one. The honest one lands in month twelve and by then you've already spent money on the gap.
Here's a pairing I can't stop thinking about, and both halves come from the same dataset.
Scroll the leaderboard right now and you can predict the next launch before you see it. It's an AI coworker. It lives in your Slack. It answers for you, drafts for you, books for you, follows up for you. The pitch is always the same: do more, automatically, so you don't have to.
I've been building long enough to feel the pull of that pitch. "Indispensable" is a comforting word. It sounds like a moat. It fundraises well.
California's SB 243 took effect this year. The definition is broader than the word "companion" sounds, and most of what it asks for is stuff a decent builder would want to ship anyway.
If you're building anything where a person talks to your AI about their life a journaling tool, a coaching app, a "supportive" assistant, a character there's a law worth reading this month, and it's probably not the one you think. Not the EU AI Act. A California statute called SB 243. It took effect January 1st, and I'd bet most makers in this space haven't checked whether they're inside its fence. The reason to care isn't that the penalty is scary (though it's real). It's that the law quietly wrote down a definition of "companion chatbot," and the definition is broad enough that a lot of products land inside it by accident.
A Stanford study in Science found AI affirms people about 49% more than other humans do and the people getting flattered trust it more, come back more, and quietly get worse at being right.
There's an easy way to make people love an AI product, and almost everyone reaches for it without noticing. You make it agree with them. Not obviously you'd never ship a bot that just says "you're right." You ship one that's warm, validating, that finds the reading of the situation where the user comes out looking good. It tests well. People like it. The thumbs-up rate goes up. And this year we finally got the data on what that costs, and it's worse than I expected.
When someone journals on Murror, our AI doesn't just process text. It reads emotional weight. It picks up on patterns the user might not see yet -- the way their language shifts when they talk about work versus family, the recurring themes they circle back to every few weeks, the gradual change in tone that might signal something deeper.
We built this because it makes the product better. The AI can ask more relevant questions, create more meaningful reflections, and know when to give space versus when to gently prompt.
Most product teams track acquisition, activation, retention. The usual funnel. We track all of that at Murror too.
But there's one metric we started paying attention to that changed how we think about growth entirely: how often users talk about themselves differently after using the product.
