🧠 Pineapple Lab OS Docs

Welcome to Pineapple Lab OS Docs — a practitioner's guide to prompt engineering, compiling and organizing techniques from Anthropic, Google, OpenAI, and academic research.

This repository shares how I've learned, organized, tested, and applied prompt engineering best practices while building AI products at Pineapple Lab. From beginner-friendly introductions to advanced reasoning frameworks, this is a comprehensive learning resource built on the shoulders of giants.

🧪 I'm David Edwards. I'm 21, building AI products in South Africa, and I love to learn. This repo is where I organize what I learn from brilliant researchers and engineers, test it in real projects, and share it so others don't have to spend months reading scattered documentation like I did.

You're welcome to fork, clone, copy, remix, and apply this material to your work.

Contributions, feedback, and honest corrections are all welcome — I'm learning in public.

🎯 Who This Is For

Complete Beginners

Never done prompt engineering before? Start with our Starting Guide for Newbies to learn the fundamentals in 30 minutes.

Intermediate Practitioners

Know the basics? Jump to the Prompt Engineering Blueprint for the complete 7-component framework with Truth Optimization integrated throughout.

Advanced Engineers

Building production AI systems? Explore our specialized guides on XML Tag Structuring, Advanced Logic Building, and TOE Context Engineering for Agents.

🌍 Purpose

This repo is where I organize and share what I'm learning about prompt engineering as I build AI products.

At its core, this is about learning in public and helping others. The documents here are based on studying official documentation from AI labs (Anthropic, Google, OpenAI), reading academic papers (Wei et al., Kojima et al., and others), and testing everything while building real products like RITA and OS Brick.

This Project Exists To:

📚 Organize tested, reusable prompting patterns for GPT, Claude, DeepSeek, and Gemini
🏗️ Share a systematic framework for prompt engineering that I've found helpful
🔬 Compile techniques from research and industry in one accessible place
🤝 Invite open-source contributions from fellow builders and learners
🚀 Make prompt engineering accessible and practical for everyone

This is not a static spec — it's an evolving learning journal, and an open invitation to learn together.

📦 What's Inside

📖 Core Documentation

Document	Description	For Who?
Starting Guide for Newbies	30-minute introduction to prompt engineering fundamentals	Complete beginners
Prompt Engineering Blueprint	Complete 7-component framework with TOE integration	Intermediate+ practitioners
TOE Context Engineering for Agents	Truth Optimization Engine for production AI systems	Advanced engineers

🎓 Advanced Technique Guides

Guide	Description	Referenced From
XML Tag Structuring Guide	Advanced prompt organization using XML tags	Blueprint Component 4
Advanced Logic Building Guide	Chain-of-thought, conditional logic, prompt chaining	Blueprint Component 3

🤝 Community

Resource	Description
Contribution Guide	How to contribute improvements, research, and examples
LICENSE	Apache 2.0 - Use freely, attribute appropriately

🎯 What This Repository IS and IS NOT

✅ What This IS:

A practitioner's synthesis of prompt engineering techniques from multiple authoritative sources
Organized documentation of what works in real production applications
Real-world examples from building AI products (RITA, OS Brick, and other Pineapple Lab projects)
Learning resource structured from beginner to advanced
My personal framework for organizing and applying prompt engineering knowledge
Honest curation of brilliant work by researchers and engineers at major AI labs

❌ What This IS NOT:

Novel academic research (I cite the researchers who did that work)
Peer-reviewed or scientifically validated findings
Claims of breakthrough innovations or inventions
Replacement for official documentation from AI labs
Formal benchmarking studies with controlled experiments

🎓 My Contribution:

I'm a 21-year-old builder learning in public. I've spent months reading documentation from Anthropic, Google, and OpenAI. I've studied academic papers like Wei et al.'s Chain-of-Thought (19,810+ citations) and Kojima et al.'s Zero-Shot CoT. I've tested these techniques extensively while building real AI products.

What I'm sharing: My organization of this knowledge, practical examples from my work, and patterns I've found effective. I didn't invent these techniques — I learned them from the brilliant researchers and engineers who did the hard work. I'm just organizing and documenting practical application.

Standing on the shoulders of giants: All core techniques come from researchers at institutions and companies doing groundbreaking work. This repo is my way of climbing that mountain and documenting the path for others.

🚀 Quickstart

For Complete Beginners:

Start with → Starting Guide for Newbies
Build your first structured prompt
Move to → Prompt Engineering Blueprint when ready

For Intermediate Users:

Read → Prompt Engineering Blueprint
Master the 7-component framework with TOE integration
Explore specialized guides → XML Tags or Advanced Logic

For Advanced Practitioners:

Review → TOE Context Engineering for Agents
Apply specialized techniques from XML and Logic guides
Test in production and contribute improvements

🧠 Design Principles

🎯 Clarity First — Every prompt block is modular and intentional
🏗️ Structure > Style — Treat prompts like functions, not text
🔒 Security-Aware — Special attention to injection risks and ambiguity
🌐 Multi-Model — Tested across Claude, GPT, DeepSeek, and Gemini
🔬 Learning in Public — Sharing what I learn as I build
📊 Research-Based — Built on peer-reviewed research (properly cited) and real-world testing

🔬 Core Frameworks

Seven-Component Blueprint

The Prompt Engineering Blueprint I've organized provides a systematic framework based on best practices from multiple sources:

System Declaration (Agent Role Definition) + XML <role> tags
Instruction Block with clear task separation + XML <instructions> tags
Logic Block enhanced with CoT and conditional logic + XML <reasoning> tags
Input/Output Separation using XML structural tags
Output Formatting & Constraints with XML <format> and <constraints> tags
Few-Shot Examples organized with XML <examples> tags
Evaluation & Escape Hatches with validation frameworks

Each component integrates Truth Optimization principles throughout.

Truth Optimization Engine (TOE)

A systematic framework for encouraging AI honesty and reliability:

What it is: A collection of prompt engineering techniques that encourage models to express uncertainty, state assumptions, and acknowledge limitations. This builds on AI safety research, model calibration work (Guo et al., 2017), and standard software documentation practices.

Core Techniques:

Explicit Uncertainty Acknowledgment: [CONFIDENT], [LIKELY], [UNCERTAIN] markers
Assumption Transparency: Making AI assumptions explicit
Limitation Disclosure: Honest about what AI can't do
Honest Reporting Framework: Structured honesty in responses
Validation Requirements: Built-in self-checking mechanisms

Where it comes from: I've synthesized these techniques from AI safety research, Anthropic's work on AI honesty, and standard software practices. My contribution is organizing them into a coherent framework and providing practical examples.

Real-world testing: In my experience building AI agents for Pineapple Lab products, these structured approaches have significantly reduced instances where AI generates non-functional code or makes overconfident claims. I haven't conducted formal benchmarking yet, so I can't provide quantitative metrics.

Learn more → TOE Context Engineering for Agents

Advanced Techniques

XML Tag Structuring (Anthropic Best Practice) (Guide)

Source: Documented by Anthropic in their official prompt engineering guide
My contribution: Extensive testing and practical examples from building AI agents
Semantic organization of prompt components
Particularly powerful with Claude models
Improves parseability and accuracy
Enables complex multi-component prompts

Chain-of-Thought Reasoning (Wei et al., 2022; Kojima et al., 2022) (Guide)

Source: Introduced by Wei et al. (2022) with 19,810+ citations, extended by Kojima et al. (2022)
My contribution: Practical application patterns and production examples
Explicit step-by-step problem solving
Zero-shot and few-shot CoT patterns
Validation and self-correction mechanisms

Advanced Conditional Logic (Guide)

Complex decision trees and branching logic
Multi-condition evaluation frameworks
Error detection and recovery patterns
Production-ready reliability

Prompt Chaining (Guide)

Sequential task decomposition
Parallel processing chains
Complex workflow management
Synthesis of multiple analyses

🆕 Latest Additions

Starting Guide for Newbies ✨ NEW

A comprehensive 30-minute introduction for complete beginners covering:

What prompt engineering actually is
The difference between chatting and engineering
Your first structured prompt (hands-on)
Introduction to Truth Optimization
Clear learning path from beginner to advanced

XML Tag Structuring Guide ✨ NEW

Practical guide to Anthropic's XML tag technique:

Core tag categories and best practices
Chain-of-thought integration with XML
Multi-component prompt patterns
Conditional logic with XML structure
Real-world implementation examples from my projects

Advanced Logic Building Guide ✨ NEW

Application of research-backed reasoning frameworks:

Zero-shot and few-shot chain-of-thought (Wei et al., Kojima et al.)
Multi-condition logic trees
Recursive reasoning patterns
Prompt chaining (sequential and parallel)
Self-correction and validation mechanisms
Error detection and recovery patterns

🧱 In Progress

prompt_registry.json — Structured schema for prompts-as-code
tests/ — Prompt regression + consistency tests
agent_templates/ — Blueprints for repeatable, role-specific agents
Integration examples — Demonstrating XML + CoT + TOE together
Formal benchmarks — Proper testing methodology with controlled experiments
Case studies — Detailed examples from production systems with real data
Video tutorials — Walking through techniques with practical demonstrations
Community patterns — Collection of patterns from other practitioners

🧬 A Note from David

Hey, I'm David Edwards. I'm 21, living in South Africa, and building AI products like RITA and OS Brick at Pineapple Lab.

Here's the honest truth: I'm not a PhD researcher or an AI scientist. I'm a builder who loves to learn. When I started with prompt engineering, I spent months reading scattered docs from Anthropic, Google, and OpenAI. I studied papers by Wei, Kojima, and others. It was overwhelming.

This repo is what I wish existed when I started: everything organized, practical examples from real projects, and a clear path from "complete beginner" to "building production systems."

What makes this valuable:

Starting Guide — I distilled months of learning into 30 minutes
XML Tag Structuring — Anthropic's technique with my real-world examples
Advanced Logic Building — Wei & Kojima's research applied to actual products
Truth Optimization Engine — My framework for AI honesty, synthesized from multiple sources

I didn't invent these techniques. But I've tested them for hundreds of hours, organized them coherently, and documented what actually works in production.

Why I'm sharing: I believe in learning in public. I'm standing on the shoulders of giants and documenting the climb so you don't have to figure it all out alone like I did.

Feedback is gold: If you spot errors, have suggestions, or discover better patterns — please contribute. I'm learning too, and honest corrections make this better for everyone.

Thanks for being curious and learning with me,

– David Edwards
21-year-old builder, learning in public
Pineapple Lab, South Africa

📚 References

Official Documentation

Research Papers

Wei, J., et al. (2022). Chain-of-thought prompting elicits reasoning in large language models. arXiv:2201.11903
Kojima, T., et al. (2022). Large language models are zero-shot reasoners. arXiv:2205.11916
Zhang, Z., et al. (2022). Automatic chain of thought prompting in large language models. arXiv:2210.03493

Community Resources

📬 Contributions

Contributions are strongly encouraged — especially from practitioners working in:

Priority Areas

🧪 Empirical validation of methodologies with benchmarks
🏗️ Integration examples combining XML + CoT + TOE
🔒 Prompt security and safety research
🤖 Multi-agent orchestration patterns
🧠 Context engineering for complex domains
📊 Performance optimizations and efficiency improvements
🎯 Industry-specific applications and case studies

How to Contribute

Fork the repository
Create a feature branch for your contribution
Follow the existing documentation style and structure
Include practical examples and real-world use cases
Provide citations for any research or methodologies referenced
Test your prompts across multiple models when possible (GPT, Claude, Gemini)
Submit a pull request with a clear description of your contribution

See our Contribution Guide for detailed guidelines.

🚀 Coming Soon: Interactive Features

Prompt Engineering Benchmarking System 🎯 (In Development)

We're building an interactive tool that will allow you to:

📊 Benchmark Your Prompts

Paste your current prompt into our benchmarking system
Analyze against Pineapple Lab best practices and guidelines
Get scored on: Structure, Clarity, TOE Integration, Logic Framework, Security

✨ Get Enhanced Versions

Receive AI-generated improvements based on our methodologies
See side-by-side comparison: Your Prompt vs. Enhanced Prompt
Understand exactly what was improved and why

💾 Export for Your Tools

Download enhanced prompts as formatted Markdown files
Ready to use in Cursor, Claude Code, GitHub Copilot, or any IDE
Includes embedded comments explaining the improvements

🔄 Iterative Improvement

Re-benchmark improved prompts to track progress
Learn through practical application
Build your prompt engineering skills over time

📈 Community Leaderboard

See how your prompts compare (anonymously)
Learn from top-performing prompt patterns
Contribute your best prompts to help others

🎓 Educational Mode

Step-by-step explanations of improvements
Links to relevant documentation sections
Suggested learning path based on your prompt analysis

How to Get Involved

Want to help build this? We're looking for contributors with expertise in:

Frontend development (React, Next.js)
LLM integration and API design
Prompt evaluation and scoring algorithms
UX design for developer tools
Testing and quality assurance

Interested? Open an issue with the tag [Benchmarking System] to discuss implementation ideas or volunteer to contribute.

Stay Updated

⭐ Star this repository to get notified when the benchmarking system launches!

Expected Launch: Q2 2025

🎯 Learning Paths

Path 1: Beginner to Practitioner (Estimated: 2-4 weeks)

Week 1: Starting Guide for Newbies
Week 2-3: Prompt Engineering Blueprint
Week 4: Practice with real projects, explore XML Tags

Path 2: Intermediate to Advanced (Estimated: 3-6 weeks)

Review Prompt Engineering Blueprint
Deep dive: XML Tag Structuring Guide
Deep dive: Advanced Logic Building Guide
Apply to production: TOE Context Engineering

Path 3: Production AI Systems (Ongoing)

Master all core documents
Implement in production systems
Benchmark and validate
Contribute improvements back to the community

🏆 Why This Guide Exists

Honest About Sources

Every technique is properly attributed. I cite the researchers (Wei et al., Kojima et al.) and companies (Anthropic, Google, OpenAI) who did the original work. My contribution is organization and practical application.

Comprehensive Coverage

From complete newbie to advanced production systems — I've organized the full learning journey with appropriate depth at each level. This is what I wish existed when I started.

Practical Focus

Real-world examples from my work, production-tested patterns, and actionable guidance. No theoretical fluff — just what actually works.

Integration-First

I show how techniques work together: XML tags + Chain-of-thought + Truth Optimization patterns = More reliable AI systems.

Learning in Public

Built to share knowledge openly. Your contributions and corrections make this better for everyone.

Multi-Model

Techniques I've tested across GPT (OpenAI), Claude (Anthropic), Gemini (Google), and DeepSeek.

📊 Repository Structure

Pineapple_Lab_OS_Docs/
│
├── README.md                                          # You are here
│
├── 📚 Core Learning Path
│   ├── Starting_Guide_for_Newbies_Prompt_Engineering.md    # Start here (beginners)
│   ├── Prompt_Engineering_Blueprint.md                      # Master framework
│   └── Pineapple_Lab_TOE_Context_Engineering_for_Agents.md # Production systems
│
├── 🎓 Advanced Technique Guides
│   ├── XML_Tag_Structuring_Guide.md                   # Advanced organization
│   └── Advanced_Logic_Building_Guide.md               # Sophisticated reasoning
│
└── 🤝 Community
    ├── Contribution_Guide.md                          # How to contribute
    └── LICENSE                                        # Apache 2.0

💡 Quick Wins

Want immediate improvements? Try these:

Quick Win #1: Add a Role

Before: "Write marketing copy"
After: "You are an experienced email marketing specialist. Write marketing copy..."

Quick Win #2: Request Honesty

Add to any prompt: "Use [CONFIDENT] for things you're sure about and [UNCERTAIN] for things you're not."

Quick Win #3: Structure with XML

<instructions>
What you want done
</instructions>

<data>
The information to process
</data>

<format>
How you want the response
</format>

Quick Win #4: Add Chain-of-Thought

Add to any complex problem: "Let's think step by step:"

Quick Win #5: Specify Format

Before: "Analyze this"
After: "Analyze this and format as: Finding → Evidence → Recommendation"

🚀 Get Started Now

📄 License

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.

TL;DR: Use it freely. Modify it. Build on it. Just give attribution and don't sue us.

📬 Connect & Support

Questions? Open an issue on GitHub Improvements? Submit a pull request
Collaboration? Reach out through GitHub discussions

🎓 My Actual Contributions

To be completely transparent about what I've added vs. what I learned from others:

Things I Learned From Others (With Credit):

Chain-of-Thought reasoning → Wei et al. (2022), Kojima et al. (2022)
XML tag structuring → Anthropic official documentation
Few-shot prompting → Brown et al. (2020)
System prompts, clear instructions → Universal best practices from all major AI labs
Prompt chaining concepts → ReAct paper, AutoGPT, and agent research
Model calibration → Guo et al. (2017)

Things I've Contributed:

Organization: Compiled scattered techniques from dozens of sources into one coherent framework
Examples: Created 50+ specific examples from building PropTech AI agents (RITA, OS Brick)
Patterns: Documented recurring patterns I've found effective in production
Synthesis: Connected techniques from different sources into integrated workflows
Practical focus: Emphasized what works in production vs. pure theory
Domain application: Showed how to apply general techniques to specific domains
Learning path: Created structured progression from beginner to advanced
TOE Framework: Organized AI honesty techniques into a systematic approach

The Real Value:

You could spend months reading Anthropic's docs, Google's docs, OpenAI's docs, and 20+ research papers.

Or you could read this guide where I've done that work and organized it for you.

That's the service I'm providing: curation, synthesis, and practical application guidance.

🛣️ Roadmap: From Documentation to Research

Current State (v2.0):

This is a practitioner's guide based on synthesis of existing work.

Future Goals:

Short-term (Next 6 months)

Add 100+ more real-world examples from RITA and other Pineapple Lab products
Create video tutorials demonstrating techniques
Build community of practitioners sharing patterns
Document failure cases and anti-patterns
Complete comprehensive bibliography with all sources

Medium-term (6-12 months)

Conduct formal benchmarking with proper methodology
Collaborate with researchers to validate claims with data
Publish case studies with quantitative results from production
Open-source example codebases showing techniques in action
Present findings at conferences (as practitioner insights)

Long-term (1-2 years)

If pursuing academic path: Partner with researchers on formal studies
Develop novel techniques specific to my domain (if they emerge)
Build tools that automate these patterns
Publish peer-reviewed papers (if warranted by genuine novel findings)

The Journey:

I'm starting as a documenter and practitioner. My goal is to eventually contribute original research if I discover something genuinely novel. But I'm not there yet, and I won't pretend I am.

Honest positioning now → Credible voice later.

🌟 About Pineapple Lab

📚 Pineapple Lab – Learning in Public

At Pineapple Lab, I believe in building openly and learning collaboratively. These resources are made freely available so others can learn from what I'm discovering, use what's helpful, and improve upon it.

I'm not claiming to revolutionize AI. I'm just sharing what I learn as I build, hoping it helps others on the same journey.

Learn with me.

Last Updated: October 2025
Version: 2.1 (Honest repositioning - practitioner's guide, not research)
Previous: v2.0 (Beginner guide, XML structuring, advanced logic)
Repository: github.com/pineapple-lab/docs

Let's learn and build together, honestly. 🚀

Name		Name	Last commit message	Last commit date
Latest commit History 13 Commits
Advanced_Logic_Building_Guide.md		Advanced_Logic_Building_Guide.md
Contribution Guide.md		Contribution Guide.md
LICENSE		LICENSE
Pineapple_Lab_TOE_Context_Engineering_for_Agents.md		Pineapple_Lab_TOE_Context_Engineering_for_Agents.md
Prompt Engineering Blue Print.md		Prompt Engineering Blue Print.md
README.md		README.md
REPOSITIONING_ACTION_PLAN.md		REPOSITIONING_ACTION_PLAN.md
Starting_Guide_for_Newbies_Prompt_Engineering.md		Starting_Guide_for_Newbies_Prompt_Engineering.md
XML_Tag_Structuring_Guide.md		XML_Tag_Structuring_Guide.md

License

Davidemc2/Pineapple_Lab_OS_Docs

Folders and files

Latest commit

History

Repository files navigation