MPL AI Engineering Standards

Every developer with MetaProp Labs understands these core concepts.

This site defines the baseline competencies we expect from every engineer who ships code with AI assistance. It is tool-agnostic by design: the principles transfer across any AI coding assistant, though we include practical examples referencing specific tools where it helps.

Core Philosophy

AI coding assistants are collaborators, not replacements. They are powerful when properly guided and dangerous when blindly trusted. The gap between productive AI-assisted development and expensive AI-generated chaos comes down to preparation, constraints, and verification.

Think of your AI assistant as a junior developer who knows how to code but knows nothing about your project, your business, or your standards. You would never hand a new hire a ticket and walk away. You would onboard them: explain the codebase, share the coding standards, point them to the docs, and review their pull requests carefully. The same discipline applies here.

The developers who get the most out of AI share a common trait: they invest upfront. They write clear specifications before generating code. They set up guardrails that catch mistakes automatically. They review output like a senior reviewing a junior’s PR. And they never stop experimenting with better ways to work.

The 12 Pillars

Pillar 0: LLM Foundations Understand the machinery well enough to reason about behavior.

Pillar 1: Context Engineering The quality of AI output is determined by the quality of context you provide.

Pillar 2: Planning Before Code Never let AI write code without a plan and a specification.

Pillar 3: Prompt Engineering Prompting is a core engineering skill that requires practice and intentionality.

Pillar 4: The AI as Collaborator Treat your AI assistant like a junior developer who needs onboarding.

Pillar 5: Guardrails and Quality Constraints multiply output; linting, tests, and type checking are non-negotiable.

Pillar 6: Verification and Security Never trust AI output blindly; review everything, trust nothing by default.

Pillar 7: Workflow and Tooling Be the orchestrator, not the bottleneck; leverage agents, sessions, and tools.

Pillar 8: Continuous Evolution The best AI developers never stop experimenting with how they work.

Pillar 9: Evaluation and Measurement Prompts that ship in your application are production code; test them like it.

Pillar 10: Data Hygiene and IP What you put into AI tools matters as much as what comes out.

Pillar 11: Knowing When NOT to Use AI The best AI developers know when to put it down.

Current Toolchain

Our pillars are tool-agnostic. The toolchain is our opinionated, point-in-time stack and changes as the ecosystem evolves.

Current Toolchain What we actually use today: models, agents, hooks, evals, and cost levers.

Resources

Engineering Foundations Canonical software engineering references every developer should know (non-AI-specific).

Learning Paths Curated public resources for going deeper on each pillar (AI-specific).

Reading This Guide

• New to the team: Read each pillar in order, then skim the toolchain. The pillars are intentionally concise; the learning paths take you deeper.
• Already working: Reference individual pillars when you need a refresher or want to share expectations with a teammate. Each pillar is designed to be linkable in PR reviews and onboarding conversations.
• Want to contribute: Open a PR. This site follows the same review process as our code. Suggest learning resources, sharpen pillar wording, or contribute insights from a project back to the team.

About MetaProp Labs

This site is published by MetaProp Labs. It is reviewed quarterly and licensed under CC BY 4.0.