Product Development And Production Methodology

Overview

This repository is a spec-driven, AI-native product development operating system where structured Markdown (knowledge layer) defines the system, AI agents execute against it to generate and run the product (execution layer), and data feeds back to continuously improve it—making the repository itself an executable blueprint for both digital and physical product creation.

It represents a new methodology for creating products and services as AI becomes the primary execution and production layer. AI is no longer limited to writing code; it is increasingly capable of conducting research, running tests, generating new materials, prototyping products, and operating automated production systems. As a result, new paradigms are emerging for how we communicate intent to AI. This product development OS is my approach to structuring that communication—translating strategy into executable specifications that AI can reliably interpret and act upon.

This repository serves as a reusable template that is copied each time a new product is created. It provides the structured, executable framework required for AI to understand, build, and manage the product and its production lifecycle. As the system evolves, I continuously refine and expand this repository to improve how effectively AI can operate within it.

But it is not just for products. My thoughts are to have a template that is applicable for anything, from starting a business, designing new products, launching them, and creating AI agents and workflows aginst it all. What if I want to design an agent to remove, clean and reassemble a rear hub on my road bike? Use this process. Much of the future will be run by agents doing digital and physical tasks. This OS is attempting to lay the groundwork for building agents to do things.

In this model, changes to something (company, product, service, workflow, process) are made by updating the executable structure—not manually modifying code or production systems directly. The knowledge layer defines the desired state, and AI acts as the development and production engine that implements those changes.

This framework guides the full lifecycle for starting a company and building products: how to define a product, how to build it, how to produce it, and how to continuously improve it through feedback from real-world data. While immediately applicable to software, it is equally designed for physical products, as AI increasingly integrates with automated testing, prototyping, and manufacturing systems.

I organize the layers according to the 8 core functional areas of a business that I created, as defined here.

Process Boundaries:

• If humans need to think through it → Google Docs to allow for shared working
• If AI needs to read it to build something → Markdown in a repo
• You want to map strategy >>> structured specs >>> agent execution. Strategy is business realm (problem you solve, differertiators, moats, growth strategy), which gets turned into structures specs that AI agents use to build the product to satisfy the strategy.

The most valuable stack is:

1. problem framing
2. system design
3. data modeling
4. AI orchestration

3-layer system

1. Knowledge layer (specs/) → what the system is supposed to do and how to do it. The truth via highly structured executable knowledge that AI can understand. Structure leads to intelligence. Smarter AI models help, but if poorly structured, then failure. Well structure knowledger layers can succeed even with average models.
2. Product layer (tasks/ and srv/) → takes the knowledge layer and executes it: develop code, develop the product, produce the product.
3. Data layer (data/) → what the system observes, stores, and learns from based on execution and real-world response (like from customers on product use, feedback, results, outcomes). The data validates reality, that leads to iterate above to improve.

Intelligence Vs Code

Intelligence is in specs/.

Execution is in the srv/, and execution is deterministic. Deterministic is code calls, code execution, cron jobs. It just does, but does not think. If the task is exact and repetable, use deterministic code.

If the task requires judgement or interpretation, then it is in specs/ (skill.md files to be exact). It use to be that we had to force everything into code - logic that was fuzzy, required interpretation and still terrible at intelligence. Now, we can collapse that cognitive logic that was in a lot of code into skills. LLMs introduce massive intelligence into the process that never existed. So, now, we have code that executes and scales, but now we can layer on intelligence to do it for us, in place of us humans.

Instead of trying to encode complex reasoning in deterministic code (if/then logic), we move that reasoning into spec/skills that guide the LLM. We keep all execution, control flow, and precision tasks in code, and only offload judgment and interpretation to the model.

Code scales execution and guarantees correctness.

Spec/Skills scale intelligence and adapt automatically as models improve. When the model improves, every skills improves automatically without changing the code. That is how we can get 10x-100x improvement from the same system, with an improved AI model.

Modular

Everything is modular, especially agents and code. We create small modules of everything so that when a better one comes along (either we create it or an open source one comes along), we can swap it out of the system.

Structure For Git Repo

• Changes and commits always tracked.
• Branch strategy: main (production), dev (active development), feature branches (experimental). Flow: feature>>dev>>main

Structure For Files

• Use consistent headings
• Keep sections predictable. If every file has a predictable structure, AI navigation improves dramatically.

repo/

• README.md \ high level overview of project; but copy out the contents in this file to the overview file to allow an AI to understand the basis for this repo and its structure for product development.
• CHANGELOG.md \
• VERSION.md \
• PROJECTRULESPROCESSES.md \ agent and human rules/processes

specs/

The executable structure layer, organized by main function and in order

• many docs could be in these folders
• decisions.md \ be sure to always have this. Why we did what we did, why we chose A over B, etc. Architecture Decision Records (ADR). provide context, decision, risks, consequences. Document using the decision layer and biases layer
• skills go here.

1_strategy_s

• vision/concept validation/strategy/mission/goals/marketing/sales/distribution/goals/target customer/constraints/success metrics/.md \ use the B2C framework.

2_systems_s

• api-spec.md - AI capabilities, focus for this project.
• architecture.md - infrastructure, platforms, code bases, tech stack, engineering design, environment \ the long-term tech stack is here
• data-model.md - how the data flows through the product, where it ends up, how it is , structure, relationships, rules governing it, where stored
• environments.md - multiple environments that exist - dev, staging, production
• system.md \ map of how the knowledge layer connects to the code, how data flows, how AI should reason about the repo
• config/
• • .env files, API's, feature flags, for both dev and production

3_product_development_s

• user-flows/screen states/actions.md
• system-ontology-framework: A framework that defines a system ontology approach where a complex software system is modeled as a structured graph (nodes and relationships) that serves as the single source of truth—allowing AI to automatically generate diagrams, detect issues, simulate changes, and even build code. It’s important because it transforms system design from static documentation into a machine-readable, AI-operable model, enabling automation, consistency, and far more intelligent development and scaling of software systems.

4_finance_s

5_production_logistics_s

6_marketing_s

7_sales-distribution_s

8_support_s

tasks-workflows

Workflows And Agents

Let's define these.

1. Execution Units is something that does work: py script, cron job, API call, function
2. Service Layer is a reusable unit that performs a defined function:
3. An agent is a system that decides what to do next. It is the harness that calls a LLM model to help apply judgement and reasoning about what to do next.

• It does not have to call an LLM - it could be a rules-based decision tree or a predefined flow. We could call this an automation controler.
• And if judgement and reasoning are required, we can call this an LLM controler, because now it calls for an LLM.

For me, a project (this repo) could encompass many agents, and agents are either automation controlers or LLM controlers, comprised of service layers (which could on their own can call an LLM), built by execution units. An agent is a workflow, and workflows can be completed by agents or humans or a mix of both.

Starting from the top, I have projects (this repo as template), workflows in the project, agents in workflows and orchestrators (humans) in workflows. And workflows can include LLM conmtrolers or automation controlers, comprised of service layers, built by execution units.

Workflows repeat, and tasks can repeat but would be defined as one-time executions for purposes of this repo.

How to build Agents. That is another process I am building and will update here in future. Closed link for now

tasks items in separate documents

date_build-plan.md \ sequences, tasks, modules, dependencies, components

srv/

The production/execution layer for AI to execute on. Code is created in files. Code is well documented and explains why the code exists, what it does, maintenance tips, gotchas if any.

digital

digital/software-based products/services

• app/
• models/
• scripts/
• services/
• ui/
• utils/
• routes/

physical

physical-based products/services

data-ops/

*Data that is already a part of the system (ie.: CRM customer data) or which gets produced from the system. repo/ topline and docs/ are strategy, systems and product development are src/, data is all other functions, pulled from B2C Framework

1_strategy_d

2_systems_d

• json-db/
• schemas/
• runtime-exports/
• deployment.md \
• monitoring.md \
• maintenance.md \
• backup.md \
• security.md \

3_product_development_d

4_finance_d

5_production-logistics_d

6_marketing_d

7_sales-distribution_d

8_support_d

tests/

Run time feedback

• acceptance-tests.md \
• qa-checklists.md \
• dev-tests/