AIMM Hub

essay

Part 5a: An Ambient Folder, End to End (Stages 1-2)

Now we build something real. This piece and the next three walk through a complete ambient folder system—from empty directory to mature infrastructure. You’ll see how each concept from the previous pieces actually comes together in practice.

An “ambient folder” is a working directory that contains:

  • Your raw materials (documents, notes, research)
  • Your infrastructure (templates, schemas, standards)
  • Your operations (scripts, workflows, integrations)
  • Your outputs (essays, reports, knowledge base entries)

The folder knows how to read its own contents, understand your standards, and call on AI capabilities to extend your work. By the end, you’ll have a system where asking for output is trivial—the folder does the heavy lifting.

We’ll build this in three stages:

Stage 1 (this piece): Initialize the folder and scaffold the infrastructure Stage 2 (this piece): Set up context retrieval and basic operations Stage 3 (next piece): Add validation, composition, and publishing

Let’s begin.

Stage 1: The folder structure

Start with a blank folder. You’re building a writing/research practice. You’ll eventually publish essays and maintain a knowledge base. For now, you just need the skeleton.

my-ambient-folder/
├── source/              # raw materials (documents, notes, research)
│   └── .gitkeep
├── drafts/              # works in progress
│   └── .gitkeep
├── output/              # finished work ready to publish
│   └── .gitkeep
├── knowledge-base/      # permanent knowledge entries
│   └── .gitkeep
├── templates/           # essay template, concept template, etc.
├── schemas/             # data structure definitions
├── scripts/             # automation scripts
├── config.yml           # folder-level configuration
└── README.md            # folder documentation

This is the skeleton. Every folder is different, but the principles are the same:

  • source/: your raw material (research, notes, PDFs, whatever you work with)
  • drafts/: intermediate outputs while things are being shaped
  • output/: finished work ready to move to the next stage (publishing, knowledge base, etc.)
  • infrastructure: templates, schemas, scripts, config

This structure reflects the workflow: source → draft → output. The ambient layer will orchestrate work flowing through these stages.

Stage 2: Scaffold the infrastructure

Now you need to define what “right” looks like for your folder. This is where you encode your standards.

2.1 Essay template

Create templates/essay.md:

---
title: ""
description: ""
pubDate: 2026-01-01
tags: []
draft: true
---

[Body content here]

---

## References

[List sources, links, and related pieces]

This template defines what a finished essay looks like in your folder. The frontmatter is the contract: every essay has these fields. The body is flexible. The references section is where you log your sources.

2.2 Concept template

Create templates/concept.md:

---
title: ""
description: ""
kind: concept
status: draft
tags: []
created: 2026-01-01
updated: 2026-01-01
---

## Definition

[One-sentence definition]

## Core idea

[2-3 paragraphs explaining the core idea]

## Applications

[Where and how this concept shows up in your work]

## Related concepts

[Links to related ideas]

## Sources

[Where this idea comes from]

Notice: different structure than an essay. This defines the “concept” type of output. Your harness will know to use this template when creating a concept, and a different template when creating an essay.

2.3 Schema definitions

Create schemas/essay.yml:

type: essay
required_fields:
  - title
  - description
  - pubDate
  - tags
  - draft
optional_fields:
  - series
  - seriesOrder
validations:
  title:
    type: string
    min_length: 5
    max_length: 120
  description:
    type: string
    min_length: 10
    max_length: 200
  tags:
    type: array
    allowed_tags: [ai-era, writing, agents, infrastructure]
  draft:
    type: boolean

This schema is the validation contract. When the harness creates an essay, it checks the output against this schema. Does the title exist? Is it between 5 and 120 characters? Are the tags from the allowed list?

If validation fails, the harness knows something went wrong and can either retry or alert you.

Create schemas/concept.yml similarly, with the fields and validation rules specific to concepts.

2.4 Configuration

Create config.yml:

folder:
  name: "My Ambient Practice"
  description: "A personal knowledge and writing practice"
  
defaults:
  output_format: markdown
  language: English
  author: "Your Name"
  
standards:
  essay_tags: [ai-era, writing, agents, infrastructure, methodology]
  concept_status_values: [draft, review, published, archived]
  date_format: YYYY-MM-DD
  
ai_context:
  voice_style: "analytical, precise, vivid"
  audience: "practitioners building their own systems"
  output_length_essay: "2000-3000 words"
  output_length_concept: "300-500 words"

This configuration file tells the harness (and the AI) everything it needs to know about your folder. It’s the layer that says “when you generate output for this folder, here’s what success looks like.”

2.5 A master context document

Create config/context.md:

# Ambient Folder Context

## Practice Overview

This is a writing and research practice focused on [your focus area]. 
We generate essays and maintain a knowledge base of concepts and patterns.

## Standards

### Voice
- Analytical but vivid. Show don't tell.
- Specific over general. Concrete examples matter.
- Always take a point of view. Never hedged or measured.

### Structure (Essays)
- Open with a specific observation or problem
- Build the argument methodically
- Close with implications or next steps
- 2000-3000 words

### Structure (Concepts)
- Definition first
- Core idea (2-3 paragraphs)
- Where it shows up in our work
- Related concepts
- 300-500 words

### Quality gates
- All claims either supported by evidence or explicitly argued
- All outputs match the relevant template
- All essays tagged with at least one category
- No spelling or grammar errors

## Related Resources

- Knowledge base: [location]
- Publishing platform: [location]
- Backlog of ideas: [location]

This is context the harness will send to the AI. It’s not a template. It’s a briefing. It tells the AI what voice to use, what structure to follow, what quality standards to meet.

Stage 2 complete: You now have infrastructure

At this point, you have:

  • A folder structure that reflects your workflow
  • Templates that define what finished work looks like
  • Schemas that define validation rules
  • Configuration that tells the harness about your standards
  • Context documents that brief the AI

This is the foundation. You haven’t automated anything yet. You haven’t called an AI. But you’ve made your standards explicit and machine-readable.

This is stage 1 and 2 together: you’ve initialized the folder and scaffolded the infrastructure.

In the next piece, we’ll add the operations layer: scripts and integrations that make the folder actually work.