Poke.com: A Glimpse of How We'll Actually Talk to AI

  • I recently tried Poke.com, and for the first time, an AI product felt like it understood how human conversation actually works. The onboarding was remarkably smooth, but more importantly, it nailed something most AI products get wrong: relevance. Instead of awkwardly surfacing random LinkedIn details from years ago like other chatbots do, Poke.com actually surfaced things about me that mattered and were currently relevant.

  • The interaction design feels fundamentally different. You can speak to it in short, natural messages like you would a human. It responds with equally concise replies rather than verbose paragraphs. The red message indicators and typing animations eliminate that dead time where you're wondering if anything is happening. During onboarding, I found myself genuinely talking to it like a real person. The combination of short messages, read receipts, actual knowledge about me, and a distinct personality created something that felt more like the future of AI interaction than anything I've used before.

  • The personality itself is interesting, though not necessarily appealing long-term. Poke.com essentially does the opposite of recent LLMs by being somewhat abrasive rather than overly helpful and deferential. It basically negs you. While this creates a memorable personality, it's not something you'd want permanently. Fortunately, the personality does evolve after onboarding. What matters more is that it has an identity tied to a phone number that you can add to iMessage, SMS, or WhatsApp. This approach of giving AI a fixed identity that you can integrate into existing communication channels feels like the right direction. I'd love to see this extended to email addresses I could CC or Slack users I could add to channels.

  • The limitations are real though. It gets confused in group chats and you can't use it across multiple messaging platforms simultaneously. The core functionality after signup still feels incomplete, and many features I'd want aren't available yet.

  • But here's what struck me most: this was the first time I could see a realistic path to a billion-dollar single-person company. The thing actually sold me a $50 monthly subscription, and throughout the process I kept thinking "this feels like a real salesperson." If it can sell to me without any human involvement, it can probably sell to many others. That's a business model that actually scales.


  • Website Page