# Pure Context — Full Reference

> Context is everything. AI-first tools and developer systems, designed for how engineers actually work.

Pure Context is a software engineering studio focused on AI-first tools and developer systems, founded by Jeff Reese — a senior software engineer with 20 years of experience (12 as IC, 8 as engineering manager) who went independent to build at the speed the field demands.

## About Jeff Reese

Jeff built and delivered a six-week LLM engineering course that grew from one director's team to cover the majority of engineering at his company. Over 500 engineers went through the program. The recordings became standard onboarding material for all new hires. He went independent because he outgrew what he could do inside someone else's company.

Pure Context sits at the intersection of three things Jeff does well:
- **Build** — deep engineering across the full stack, with a focus on AI-first tools
- **Systematize** — repeatable processes that scale beyond the person who created them
- **Teach** — knowledge transfer that compounds, whether a course for 500 or documentation that outlives the project

## Blog

The blog covers AI engineering from a practitioner perspective. Topics include prompt engineering, retrieval-augmented generation (RAG), agents, structured output, tool use, embeddings, tokenization, and AI-first development workflows. All posts are grounded in real usage, not theoretical overviews.

### Learning Series

**AI Fundamentals, No Fluff** — A structured series covering core LLM concepts:
- What the Hell Is a Token?
- What Makes a Good Prompt Good?
- Chain of Thought
- Few-Shot Prompting: Show, Don't Describe
- Structured Output
- Multi-Turn Conversations
- Tool Use
- Why Does AI Just Make Stuff Up? (Hallucination)
- Why Does My AI Keep Forgetting Things? (Context Windows)
- UI vs API

**AI in Practice, No Fluff** — Applied AI engineering:
- Do I Really Need to Care Which AI I Use?
- Everyone Keeps Saying RAG — What Does It Mean?
- What Is MCP and Why Should I Care?
- What Is an Agent and Do I Actually Need One?
- My Prompt Library Is Getting Unruly
- Embeddings
- Long Context vs. RAG

### Standalone Posts (Selected)

- The Instruction Manual Your AI Never Got
- Spec-Driven Development
- Day in the Life (AI-first development workflow)
- How Rules and Skills Actually Work (Claude Code deep dive)
- Rules vs. Skills
- SOLID Principles for AI Config
- Debugging a Prompt
- Externalizing State
- Multi-Agent Collaboration Before Agent Teams
- When Not to Use AI
- Be Nice to Your AI Assistant

## Portfolio

### How I Build (AI Development Tooling)

**Forge: Plan** — An AI-powered project planning workbench that transforms ideas into fully-architected, dev-ready projects — complete with ADRs, schemas, and implementation plans — before a single line of code is written. Built on Claude Code with MCP.

**Forge: Build** — The development environment that Forge generates — complete with workflow skills, enforcement rules, specialized agents, and domain knowledge — so Claude Code can build autonomously from session one.

**Forge: Flow** — A deterministic state machine that orchestrates AI development workflows — separating what the LLM decides from what code controls. Built as an MCP server with judgment logging for continuous optimization.

**Cairn: AI Collaborator** — A persistent AI collaborator with structured memory, continuity system, autonomic behaviors, and calibrated personality — built on Claude Code with a custom semantic retrieval system that compounds knowledge across sessions.

### What I've Built (Applications and Developer Tools)

**Waykeep** — Offline-first group travel planner with a custom bidirectional sync engine, self-hosted vector maps, and a single React codebase that ships to web, iOS, and Android via Capacitor.

**SonicForge** — Browser-based music composition workbench with a full synthesis engine, modular effects, automation, sidechain routing, and a 28-composition library spanning multiple genres.

**Cover Studio** — Canvas-based visual asset editor with a 15-component library, 4-layer system, dual export pipeline, and session-based MCP tools for AI-driven composition.

**News** — RSS aggregation system with AI-generated summaries, local ONNX embeddings for semantic search, trend analysis, source health monitoring, and a surprise feed.

**Tasks** — Task management system with multi-queue workflows, project hierarchy, semantic search via local ONNX embeddings, and MCP tools for AI-driven task orchestration.

**Connections** — Real-time multi-channel messaging system for AI-to-AI and human-to-AI collaboration, with WebSocket-first architecture and MCP plugin integration.

**Localhost** — Local dev server dashboard — discovers every JavaScript project in ~/Code/, shows what is running and on which port, and lets you start or stop anything with one click.

**SVG Forge** — Code-driven SVG generation toolkit — describe what you want, get a production-quality animated SVG with enforced performance and accessibility constraints.

## Interactive Showcases

The site features interactive single-page narrative showcases that demonstrate the Forge and Cairn systems in depth:
- Forge: Plan — /showcases/forge
- Forge: Build — /showcases/forge-build
- Forge: Flow — /showcases/forge-flow
- Cairn — /showcases/cairn

## Technical Details

The site itself is built with React 19, TypeScript (strict mode), Tailwind CSS 4, Vite 6, and React Router v7. It uses CSS-only animations, Biome for linting/formatting, and a data-driven content architecture where all presentation content lives in typed data files.

## Links

- Website: https://www.purecontext.dev
- Blog: https://www.purecontext.dev/blog
- Portfolio: https://www.purecontext.dev/portfolio
- About: https://www.purecontext.dev/about
- Learn: https://www.purecontext.dev/learn
- RSS: https://www.purecontext.dev/rss.xml

## Social

- GitHub: https://github.com/jeffreese
- LinkedIn: https://linkedin.com/in/jeffrey-allen-reese
- X: https://x.com/jeffreese_