LMAITFU

Let Me AI That For You

A playful homage to the classic LMGTFY (Let Me Google That For You), but for the AI era.

What It Does

1. You enter a question someone asked you (that they could have easily asked AI)
2. You get a shareable link
3. When they open it, they watch an animation of:
   • The question being typed into an AI chat interface
   • The AI "thinking"
   • The actual AI response appearing

The passive-aggressive way to tell someone: "You know you could have just asked AI, right?"

How It Works

This is a static site that runs entirely in the browser. When viewing a shared link:

1. Choose an AI provider:

   • ✨ Free AI — No API key needed! Just works. (Powered by Pollinations.ai)
   • OpenAI — Use your own API key for GPT-4
   • Anthropic — Use your own API key for Claude

2. Watch the animation play while the real API call happens

3. See the actual AI response

Why Client-Side?

• No backend to maintain
• Free option via Hugging Face Inference API
• Your key (if used) = your cost
• Privacy: queries never touch a third-party server (except the AI provider)

Usage

Creating a Link

1. Visit the site
2. Type the question
3. Click "Generate Link"
4. Share the link with the person who should have asked AI themselves

Viewing a Link

1. Open the shared link
2. Click "Show Me!" (or use your own OpenAI/Anthropic key)
3. Watch the animation
4. See the AI's answer
5. Feel appropriately chastised

Technical Details

• Pure static site (HTML, CSS, JavaScript)
• Hosted on GitHub Pages
• No backend, no database, no cookies (except localStorage for API key)
• Query is encoded in the URL (base64)
• API call happens on the viewer's browser using the creator's embedded (encrypted) key hint or viewer's own key

Local Development

# Clone the repo
git clone https://github.com/davidthingsbot/LMAITFU.git
cd LMAITFU

# Serve locally (any static server works)
python -m http.server 8000
# or
npx serve .

# Open http://localhost:8000

Deployment

The site is automatically deployed via GitHub Pages from the main branch.

Live at: https://davidthingsbot.github.io/LMAITFU/

Privacy & Security

• API keys are stored in browser localStorage only
• Keys are never transmitted to any server except the AI provider
• Shared links contain the query but NOT your API key
• Recipients need their own API key to see responses (or you can include a one-time key hint)

Contributing

PRs welcome! Ideas:

• More AI providers (Gemini, local models via Ollama)
• Better animations
• Snarkier messages
• Mobile improvements

License

MIT — Do whatever you want with it.

Credits

Inspired by LMGTFY, which has been teaching people to Google since 2008.

---

For when "just ask ChatGPT" is too polite.

---

Appendix: How This Was Built

The Original Prompt

Create a new repo in work called LMAITFU. It's going to be a public repo. And it's going to be a static website. And it's going to be modeled after LMGTFU. The famous site that just animates asking Google a query. This animates asking AI query and presents what AI returns. Technically this relies on the website being able to access open AI or Claude itself. Write this up in a read me. Set it up as a Google pages repo. Add commit push.

What Happened Next

From that single voice-transcribed prompt, an AI assistant built the initial site in about 10 minutes: a dark-themed chat UI with typing animations, support for OpenAI and Anthropic APIs, and a shareable link system that encodes questions in the URL.

The first version required users to provide their own API keys. This was deemed "unacceptable for a toy app" — the whole point is frictionless passive-aggression. So began the hunt for a free, keyless AI backend.

HuggingFace was the first attempt, but their free inference API had been deprecated. Groq has a generous free tier but still requires signup. Finally, Pollinations.ai was discovered — completely free, no API key, no signup. It can be slow (10-30 seconds sometimes), but it works.

Along the way: a test suite was added (node test.js), the provider selection screen was removed for viewers (just show them the answer!), and GitHub Pages deployment was configured via Actions workflow after debugging some token permission issues.

The result: a static site where you type a question, get a link, and anyone who clicks it watches an AI answer the question they should have just asked themselves.

Total development time: ~90 minutes of human-AI conversation.