• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
AI
Reaction

Blank prompt boxes and the learning curve for consumer AI

A post argues that empty input fields leave newcomers unsure what to ask, while requests to connect personal accounts demand trust before showing what AI can do.

Josh ElmanJE
8 Sources, 13d ago, first seen 13d ago

TLDR

The writer argues that impressive AI demos, such as booking a restaurant, happen too rarely to build a habit. They propose sharing reusable task instructions rather than private results, so people can learn what to try from one another.

Combined views

41.9K

8 Sources, first seen 13d ago

461 likes114 comments450 saves42 reposts

Combined views

41.9K

8 Sources, first seen 13d ago

461 likes114 comments450 saves42 reposts
Featured Source
JEJosh Elman@joshelman9:21 AM · Sep 27, 2026
990TECH

End the Blank Box Problem Consumer AI is trying to go mainstream. But it’s too hard to learn. We have a problem. We're trying to onboard the next hundred million people to the most powerful technology we have ever built - AI and agents. But we're doing it with a blank box. You…

musical.lymusical.ly/

Sentiment

Positive44.1%55.9%Negative

Summary

Some accounts welcomed the focus on replacing the blank text box to help mainstream AI adoption through better discovery and idea feeds, while others argued the problem runs deeper since most use cases remain irrelevant to normal users.

Based on 73 sentiment-bearing replies from 61 accounts across 2 conversations.

Related

VC Josh Elman Praises Blog Post Redefining AI Inference Beyond Tokens
The ‘invisible AI layer’ behind work inside companies

A user says companies get the results of employees’ individual AI setups, but little of what makes those results good is shared.

An X chat is planned to test product-marketing AI Skills with 15–20 practitioners

The organizer says the first Skills cover launch readiness, positioning, pricing and market briefings, and work in AI tools people already use.

Sentiment

Positive44.1%55.9%Negative

Summary

Some accounts welcomed the focus on replacing the blank text box to help mainstream AI adoption through better discovery and idea feeds, while others argued the problem runs deeper since most use cases remain irrelevant to normal users.

Based on 73 sentiment-bearing replies from 61 accounts across 2 conversations.

8 Sources

Josh Elman@joshelmanEnd the Blank Box Problem Consumer AI is trying to go mainstream. But it’s too hard to learn. We have a problem. We're trying to onboard the next hundred million people to the most powerful technology we have ever built - AI and agents. But we're doing it with a blank box. You know the box. Open almost any top AI product today and you get an empty input field, maybe two or three example prompts tucked underneath. The newest wave of assistants built into your messaging apps go even further: text anything, get it done. But you have to know what to say. The most onboarding you see in these products is a giant connector screen. It feels like homework — connect your life. Your email. Your calendar. Your social accounts. Your finances. Your health. And it is a very personal ask, challenging your privacy and trust before you even know what you will get. Here's the thing. That experience is incredible — if you already know what to do. Show up with something top of mind and it's magic. AI does something for you in ways you never thought possible. I couldn’t figure out a domain redirect, and I handed my account to an agent and it fixed everything. My daughter wanted sheet music converted so that instead of repeats and codas it was a single stream of music she could play. My Muse (Alfalfa) fixed it in two shots, what used to take me 45 minutes of PDF editing. But most people don't show up with something top of mind. Most people show up curious, a little unsure, wanting to understand what this thing even is. A blank box tells them nothing. It's the worst onboarding screen ever designed, and we've made it the front door of the entire category. So people do the natural thing. They try the demos they've heard about. Book a restaurant reservation. Shop for cheaper insurance. Cancel a subscription they forgot they had. And it works — genuinely magical, the first time. And then… nothing. Because look at what those tasks have in common: you do them almost never. You book the special-occasion dinner a few times a year. You shop insurance once in a blue moon. You cancel a subscription and then there's nothing left to cancel. These are the greatest-hits demos of AI precisely because they're impressive and rare. But rare is the whole problem. A product you use twice a year isn't a habit. It's a party trick. I've written before that every product has a purpose, a set of core actions, and a cycle — the frequency of those actions. The coolest demos right now are all cycle-of-a-few-times-a-year. No daily loop, so no reason to come back, so no learning, so the magic quietly fades. So what's actually missing? Discovery. And learning from other people. Here's something we've known in consumer for twenty years and somehow forgot the moment we started building AI: people learn what to do by watching other people do it. It's the entire engine of social. The magic of http://musical.ly, and then TikTok, was never just the videos — it was that you saw someone do something, thought I could do that, and then went and did it. You didn't arrive knowing what to make. You learned by example, at scale, every single day. Good ideas spread person to person, and each one taught you a new capability of the product without a single tutorial. Now look at today's AI products. Where's that loop? You're alone with your blank box. No feed of what's possible. No sense of what people like you are getting done. No way to see a great idea and just copy it. We built the most capable products in history and then skipped the learning loops. Now I know a lot of the best AI and agentic use cases are private and personal while social media was default for sharing. But that’s no excuse. You don't need to see my daughter's sheet music to copy what I did with it. What travels is the recipe, not the result: 'turn repeats and codas into one continuous score' is shareable. Nobody's life leaks with it.13d
Hiten Shah@hnshahThe blank box is really a dead end for learning. If every good use case stays trapped inside the conversation where someone discovered it, the product never gets much easier for the next person. Feels like the next big consumer AI loop is people teaching each other what these things are actually good for.13d
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    Josh ElmanHiten ShahScott Belsky

    8 Sources

    Josh Elman@joshelmanEnd the Blank Box Problem Consumer AI is trying to go mainstream. But it’s too hard to learn. We have a problem. We're trying to onboard the next hundred million people to the most powerful technology we have ever built - AI and agents. But we're doing it with a blank box. You know the box. Open almost any top AI product today and you get an empty input field, maybe two or three example prompts tucked underneath. The newest wave of assistants built into your messaging apps go even further: text anything, get it done. But you have to know what to say. The most onboarding you see in these products is a giant connector screen. It feels like homework — connect your life. Your email. Your calendar. Your social accounts. Your finances. Your health. And it is a very personal ask, challenging your privacy and trust before you even know what you will get. Here's the thing. That experience is incredible — if you already know what to do. Show up with something top of mind and it's magic. AI does something for you in ways you never thought possible. I couldn’t figure out a domain redirect, and I handed my account to an agent and it fixed everything. My daughter wanted sheet music converted so that instead of repeats and codas it was a single stream of music she could play. My Muse (Alfalfa) fixed it in two shots, what used to take me 45 minutes of PDF editing. But most people don't show up with something top of mind. Most people show up curious, a little unsure, wanting to understand what this thing even is. A blank box tells them nothing. It's the worst onboarding screen ever designed, and we've made it the front door of the entire category. So people do the natural thing. They try the demos they've heard about. Book a restaurant reservation. Shop for cheaper insurance. Cancel a subscription they forgot they had. And it works — genuinely magical, the first time. And then… nothing. Because look at what those tasks have in common: you do them almost never. You book the special-occasion dinner a few times a year. You shop insurance once in a blue moon. You cancel a subscription and then there's nothing left to cancel. These are the greatest-hits demos of AI precisely because they're impressive and rare. But rare is the whole problem. A product you use twice a year isn't a habit. It's a party trick. I've written before that every product has a purpose, a set of core actions, and a cycle — the frequency of those actions. The coolest demos right now are all cycle-of-a-few-times-a-year. No daily loop, so no reason to come back, so no learning, so the magic quietly fades. So what's actually missing? Discovery. And learning from other people. Here's something we've known in consumer for twenty years and somehow forgot the moment we started building AI: people learn what to do by watching other people do it. It's the entire engine of social. The magic of http://musical.ly, and then TikTok, was never just the videos — it was that you saw someone do something, thought I could do that, and then went and did it. You didn't arrive knowing what to make. You learned by example, at scale, every single day. Good ideas spread person to person, and each one taught you a new capability of the product without a single tutorial. Now look at today's AI products. Where's that loop? You're alone with your blank box. No feed of what's possible. No sense of what people like you are getting done. No way to see a great idea and just copy it. We built the most capable products in history and then skipped the learning loops. Now I know a lot of the best AI and agentic use cases are private and personal while social media was default for sharing. But that’s no excuse. You don't need to see my daughter's sheet music to copy what I did with it. What travels is the recipe, not the result: 'turn repeats and codas into one continuous score' is shareable. Nobody's life leaks with it.13d
    Hiten Shah@hnshahThe blank box is really a dead end for learning. If every good use case stays trapped inside the conversation where someone discovered it, the product never gets much easier for the next person. Feels like the next big consumer AI loop is people teaching each other what these things are actually good for.13d
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet