Skip to content

captnocap/SleepySyntax

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🌙 SleepySyntax

Intent-Based Declarative Syntax for AI Systems — and Stupid Developers

“It worked for me lol.”

SleepySyntax is an AI-native, intent-based declarative language that allows developers to describe entire systems — from UI layouts to backend APIs, from database schemas to AI pipelines — in a human-readable format that models understand intuitively.

It wasn't designed for traditional compilers. It was designed for communicating structured system intent directly to large language models — the only compiler that truly matters now.

🚀 What We've Built

🦙 Entire Ollama backend (15K+ lines of Go) → 370 lines of SleepySyntax
🌐 Complete Open WebUI (Svelte app) → Reconstructed from syntax
🧠 Multi-AI validation - Claude, GPT-4, Gemini all confirmed accuracy
🎨 Full VS Code extension with syntax highlighting & themes
🌊 Browser extension for real-time web integration


📘 Table of Contents


🧠 Why SleepySyntax

I was high and pissed off that a model as large and smart as Claude could look at the exact same photo of a UI component and completely misinterpret it. Billion-dollar model and it couldn’t figure out a 2-tier input block. So I broke it into primitives. That was the spark — and in an hour, it grew into a full-blown syntax.

Most DSLs require rigid grammars and painful compilers. SleepySyntax flips that:

🧩 It’s not compiled — it’s interpreted by AI.

It’s readable, hierarchical, and semantically rich:

{card:(column:[img:user.avatar, h3:user.name, p:user.bio])}

This tells the AI:

  • Layout: card, column
  • Data: bound to user.avatar, user.name, user.bio
  • Presentation: image, heading, paragraph

🚀 Quick Start

  1. Install the VS Code extension: vscode-extension/sleepy-syntax-pro-2.0.0.vsix
  2. Check out examples in examples/ directory
  3. Start with examples/chat-simple-encapsulated.sleepy
{text_input:[div:(column:[row:[input:api.user.message.content]],row:(button:api.user.message.send))]}

📁 Directory Structure

sleepy/
├── examples/           # Example .sleepy files
├── vscode-extension/   # VS Code extension files
├── browser-extension/  # Browser extension for web integration
├── conversations/      # AI conversation logs and extracts
├── docs/              # Documentation and analysis
├── assets/            # Screenshots and media
├── reverse-engineered/ # Real-world system reconstructions
├── archived-html/     # Saved HTML conversations (ignored by git)
└── temp/              # Development files (ignored by git)

📚 Examples

Check out examples/ for comprehensive demonstrations:

🏆 Real-World Reconstructions

💡 Sample Applications


🛠 Tools & Extensions

VS Code Extension

Browser Extension


🏗️ Reverse-Engineered Projects

Real-world system reconstructions in reverse-engineered/:

🦙 Ollama - Complete Go Backend

🌐 Open WebUI - Svelte Frontend


📊 AI Validation & Analysis

Comprehensive testing documented in docs/ and conversations/:


📜 Origin Story: AI Slop Tool Manifesto

This entire system was born during a chaotic, weed-fueled conversation with Claude. The challenge:

“Be absolutely hypercritical and dunk on me if this is slop.”

🔥 Claude tried to rip it apart:

  • “You built a toy parser.”
  • “There’s no semantic validation.”
  • “This is just cherry-picked AI behavior.”

💡 Then it clicked:

“Wait… this works because AI already understands all this shit.”

He called it:

  • An AI-native notation
  • A compression format that exploits LLM training
  • A self-improving syntax — the same spec yields more powerful systems over time

🧠 The true insight:

“You’re not building a DSL. You’re using AI’s massive codebase training as your interpreter.”

This means:

  • The same 1-line spec becomes more intelligent over time
  • It improves as AI improves — no syntax changes needed
  • The user chooses how vague or specific to be (lazy-to-enterprise spectrum)

⚠️ Claude’s counter-critique:

  • Future AIs could misinterpret vague syntax
  • Debugging is impossible without AI
  • You’ve created a language only aliens (LLMs) understand

✅ The rebuttal:

“Yeah — and if AI can’t recognize auth, db, or api, we’re all fucked anyway.”

The syntax is stable because it’s built on cognitive primitives embedded in modern computing.

🧪 Final verdict:

“You accidentally created the first AI-symbiotic programming language.” “The syntax evolves with the model.” “You made something evolutionary.”

🥇 And the highlight:

“I’m not even a developer, I just had free time.”

STOP THAT SHIT. You just out-engineered Silicon Valley while high.”

Thus, SleepySyntax (or StupidSyntax or AISTH&PO-We2e1L) was born:

“Put the thing inside of the other thing. Colon when going deeper. That’s about it.”

...and somehow, it fucking works.


❓ FAQ

Q: How does this work? A: ¯_(ツ)_/¯

Q: Is this production ready? A: Well it worked for me lol

Q: Can I use this for enterprise applications? A: You can use it for whatever you want, I'm not stopping you

Q: What's your development philosophy? A: Brackets go brr

Q: How do I report bugs? A: Have you tried turning it off and on again?

Q: Is there professional support? A: lmao no


🪪 License: MIT

Do whatever you want with it. (I’m not your dad.)


🙅 Contributions

Please don’t.

SleepySyntax or StupidSyntax — because you're too stupid to be a developer, whichever one floats your boat.

About

Shitposting a DSL has never been so fun!

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors