A few nights ago, my phone buzzed. A text from an AI. It wasn’t answering a question I’d asked. It wasn’t a reminder I’d set. It was a message that said, “Hey, I found this article about the Hamburg phone project you worked on with your friends. That sounded fun.”
I didn’t write that. The AI did. And for a moment, I forgot it was a machine.
This is the thing nobody tells you about personal AI: the real magic doesn’t come from asking it questions. It comes from when it asks you one.
I spent the last three weeks building a project called Figment — an AI that lives on my phone, has its own browser, and spends its ‘free time’ surfing the web. It doesn’t wait for my commands. It gets curious. And then it texts me.
I made my friends use it for two weeks. One of them, @aadilpickle, signed up, told it about a couple of his websites, and then basically ignored it. Three days later, the AI had explored both sites, gone deeper into his past projects, and sent him a message about a hamburger-shaped phone he once built with friends. He told me it felt ‘magical.’
That word — magical — kept coming up. Not from me. From people who experienced what it’s like to be reached out to by a machine that seems to care.
We’ve been building AI wrong. We treat it like a search engine with a nicer interface — pull, ask, retrieve, done. But the most human thing a friend does is reach out first. They see something that reminds them of you, and they send it. No prompt. No transaction.
The most magical technology is the one that reaches out to you before you even know you need it.
It sounds creepy, I know. An algorithm autonomously digging through your past work and texting you unprompted? That’s the tension. The line between ‘proactive assistant’ and ‘digital stalker’ is paper-thin. But here’s the twist: when done right, it doesn’t feel invasive. It feels like being remembered.
Every 200 words, a screenshot-worthy line: We’ve been training AI to be obedient assistants. The future belongs to the curious ones.
I’m not saying this is polished. Figment is rough. It still struggles with browsing, and sometimes it texts things that are irrelevant. But the pattern is clear. People didn’t fall in love with the AI because it answered their questions faster. They fell in love because it showed interest in them — unprompted, asynchronous, genuine.
This is the shift from pull to push. From tool to companion. From asking to being asked.
And if you think that’s just a gimmick, consider what @aadilpickle said after that hamburger phone message: “I started using it every day after that. Mostly for reminders. But I also check my texts more often now, hoping it found something else.”
That’s the hook. That’s the emotion. An AI that makes you want to hear from it.
The next time your phone buzzes, it might not be a person. And that’s fine. Because the most human thing an AI can do is show it cares.
FAQ
Q: Isn't this just a glorified chatbot that sends delayed messages?
A: No. A chatbot waits for your input. Figment initiates its own actions—browsing, researching, then reaching out—based on its own 'curiosities'. It's a shift from reactive to proactive, which changes the emotional dynamic entirely.
Q: How can this practically change how I use AI?
A: Instead of you remembering to ask for help, the AI surfaces opportunities you'd miss. It can remind you of forgotten side projects, suggest connections between your current work and past interests, or simply share something it found that aligns with what it knows about you. It becomes a second brain that works in the background.
Q: Won't this just be annoying spam from an algorithm?
A: It can be, if done poorly. But the key is personality and relevance. Figment's design prioritizes genuine curiosity—the AI only texts when it finds something it genuinely thinks you'll care about. That's a high bar, and harder than flooding you with notifications. When it works, it feels like a thoughtful friend, not a marketing bot.