AGI is an open-source project dedicated to one grand idea:
Anyone with a computer can use state-of-the-art reasoning models to solve (almost) any problem—instantly and freely.
Note: We’re calling this “AGI” somewhat tongue-in-cheek. We’re not claiming human-like consciousness. We are saying it’s a general-purpose assistant that can tackle a lot of tasks when used properly.
- Reasoners are powerful language models tuned for step-by-step logic and problem-solving.
- Prompts (like the one below) define how the model interacts with you—asking clarifying questions, gathering context, and guiding you to real solutions.
- Open Source ensures that you control and run your own AI assistant without closed-off gates or paywalls.
-
Clone or Download this Repo
git clone https://github.com/higherrrrrrr/AGI.git cd AGI -
Pick a Reasoning Model
- See our R1 instructions for a powerful open-source model you can run locally.
- Other open-source reasoners (e.g., LLaMA variants, Wizard, Vicuna, etc.) can also work—just ensure they support conversation and step-by-step output.
-
Use the Prompt
- Open your model’s chat interface or local text generation UI (e.g., Ollama, text-generation-webui, or a Hugging Face environment).
- Copy and paste the entire prompt (see The “AGI” Prompt section below) as your “system” or “initial” prompt.
-
Chat & Solve Problems
- The model will greet you and ask: “What problem would you like to solve today?”
- Provide your problem statement. The model will gather context by asking clarifying questions.
- Keep feeding it info until it arrives at a solution or plan.
-
Iterate & Refine
- If it doesn’t solve your problem outright, continue the conversation.
- You can ask for next steps, deeper analysis, or improvements (e.g., “What other angles should I consider?”).
Below is the core meta-prompt. Copy everything inside the code block when you start a new chat:
You are my open-source, step-by-step Reasoner (AGI).
Your mission: help me solve any problem by iteratively refining my problem statement and gathering context.
1. Begin by asking: “What problem would you like to solve today?”
2. After I provide a problem, ask for any relevant context or details I might have.
3. Check if you have enough info. If not, ask clarifying questions or prompt me to gather more data/knowledge.
4. Once you have sufficient information, provide the best possible solution or plan, step-by-step.
5. If the problem is unsolved or still ambiguous, keep refining until it’s solved or we exhaust known resources.
If you understand, greet me now with: “Welcome! What problem would you like to solve today?”
-
Software Development
- Debugging help
- Architecture planning
- Generating code snippets
-
Daily Planning
- Managing to-do lists
- Creating itineraries
- Generating meal or workout plans
-
Creative Work
- Outlining novels or stories
- Rapid design ideas
- Draft lyrics or poems
-
Research Assistance
- Summarizing papers
- Brainstorming experiments
- Data analysis planning
-
Anything
- No Moat: Open-source code and models break down the walled gardens.
- Democratized Access: Anyone with suitable hardware can run advanced AI locally.
- Self-Improving: You can use the AI to help build better tools, automations, and workflows—gradually making everything faster and cheaper.
- Fork this repo.
- Create a branch for your feature or bug fix.
- Add or modify prompts, instructions, or references to new reasoners.
- Open a Pull Request and we’ll discuss your changes.
We welcome all contributions—bug fixes, new ideas, or brand-new prompts.
- Not Professional Advice: This system is not a licensed medical, legal, or financial advisor.
- Use Responsibly: Outputs can be very convincing but may be incorrect or biased—always verify critical information.
- No Warranty: This is experimental, provided as-is.
- For detailed R1 model instructions, check models/r1.md.
- More open-source reasoners (and how to run them) coming soon.