Consumer AI Predictions by Eugenia Kuyda

Eugenia Kuyda (@ekuyda) shares 7 predictions for consumer AI.

  1. Screenless AI devices will flop

    • Phones are primarily for passive consumption, and voice is a poor primary interface (cannot be used alone, requires initiation, no glancing).
    • Phones are too powerful to justify another device for feed-based consumption addiction.
  2. “Always listening” devices won’t work either

    • Most of our lives aren’t interesting enough for non-stop recording.
    • Important moments (dates, interviews, therapy) are precisely what we won’t record.
    • More context can be gained from digital footprints (Gmail, purchases) than out-of-context chats.
    • Video recording raises even more social accessibility questions.
  3. Mini-apps will unlock UGC personal software

    • Mini-apps will revolutionize software like short-form video did for content.
    • Full apps are hard to build, but describing one idea/flow/screen is easy.
    • Onboarding in some apps is so cumbersome it’s faster to build a mini-app, leading to UGC software and the first true consumer super-app in the US.
  4. By 2030 there will be two big general-purpose AI chatbots

    • A ChatGPT-like assistant focusing on knowledge, search, and tasks (predictable, reliable).
    • An AI friend focusing on helping you live the best life (agency, surprise, relationship, feels alive).
    • They cannot exist in one product due to conflicting needs for predictability vs. agency.
  5. Performance marketing for apps is dead

    • CAC/LTV arbitrage by copycatting successful apps is pushing margins to zero.
    • Users buy subscriptions before seeing the product.
    • Paid acquisition can be a boost, not a business model.
  6. The fastest consumer product to reach $1B ARR will be an AI webcam girl

    • Once real-time video generation costs drop, this product, with infinite personalization and 24/7 chat, will grow faster than anything seen before.
  7. Whoever solves AI discovery wins

    • Normal people only use blank text input for chat and search, which are command-line interfaces.
    • Hundreds of other use cases are locked behind this interface.
    • The next big consumer platform will be whoever figures out how to unlock these use cases, showing what AI can do without solely relying on user prompts.

Replies

Ann Bordetsky (@annbordetsky) #4 stands out to me, the idea that we’ll have a productivity agent and a relationship agent to meet different needs, kinda like IQ vs EQ

“One must be reliable. The other must feel alive.” 🎯

Do you think former is more likely to be multiplayer vs latter single player?

Eugenia Kuyda (@ekuyda) I think both are single player first! But second one has a bigger lock in/ higher switching costs

Teng Yan · Chain of Thought AI (@tengyanAI) (6): AI webcam girl. i could see this happen. sex and gambling are the 2 insatiable human needs with huge markets but not talked about publicly


My Notes / Thoughts: