type	Manual
title	Improving Wiki
description	A unified, incrementally growing Open Knowledge Format wiki of Improving's AI thought leadership.
created	2026-10-17
updated	2026-10-17
tags	readme okf ai improving

Improving Wiki

A unified, incremental discovery wiki of Improving's AI thought leadership.

This repository gathers the thinking Improving and its leaders publish about AI — how enterprises adopt agents without losing reliability, accountability, or trust — into one cross-linked, human-readable, and agent-readable knowledge base. It is not a mirror of any single blog. It is a synthesis: every source is summarized once, then distilled into reusable concepts and entities that connect ideas across authors, years, and publications.

The wiki is built on the Open Knowledge Format (OKF) v0.1 — a plain directory tree of UTF-8 markdown files with YAML frontmatter. It needs no database, SDK, or special tooling: it reads in a text editor, renders on GitHub, diffs cleanly in version control, and is directly consumable by AI agents.

What's Inside

The bundle currently holds 1,105 source summaries and 869 cross-linked wiki pages (439 concepts, 430 entities), drawn from three blogs:

Source	Author	Posts	Focus
Emptying the Cup	Devlin Liles	354	Agentic software delivery, AI governance, enterprise retrieval, observability, real-time video, and the philosophy and economics underneath.
Driven To Develop	David O'Hara	240	Leadership, culture, personal development and health, the DFW/Texas AI economy, and the consulting business model.
Improving's Thoughts	Improving (many authors)	511	The firm's collective voice: AI in the enterprise, data engineering, platform engineering, cloud, security, delivery practice, UX, and culture.

The through-line is AI in the enterprise: how organizations move AI from experiments into measured production workflows without giving up reliability or trust. Liles supplies the engineering depth — quality gates, orchestration, evaluation, retrieval. O'Hara supplies the executive frame — governance economics, workforce restructuring, and the regional build-out. The Improving corpus adds the delivery-practice layer — how a services firm packages that thinking for clients.

Repository Layout

├── README.md          # This file
├── AGENTS.md          # The operating manual for agents maintaining the wiki
├── index.md           # Root index — the bundle's front door
├── overview.md        # High-level synthesis of themes and open questions
├── log.md             # Chronological operations log (newest first)
├── concepts/          # Common nouns: techniques, patterns, ideas
├── entities/          # Proper nouns: people, organizations, products, places
├── sources/           # One summary per ingested reference
├── queries/           # Filed question-and-answer sessions
├── references/        # READ-ONLY raw source documents, organized by origin
└── scripts/           # Reusable fetchers that populate references/

Start here: the root index lists every page. For orientation, read the overview; to see the raw material, browse references; to see how the wiki is maintained, read AGENTS.md.

How It Grows

The wiki is incremental: it expands as new writing appears, one source at a time, without rebuilding what is already there.

1. Fetch — the scripts in scripts/ archive blog posts into references/, exactly as published, and maintain the per-source indexes. They never edit the raw documents.
2. Ingest — the ingest workflow in the operating manual summarizes each source, extracts the recurring concepts and entities it introduces, and cross-links them to existing pages.
3. Curate — for a large corpus, a page is created only when a concept or entity is central to, or recurs across, two or more sources. One-off mentions stay in the source summary's prose until a later source makes them worth a page. This keeps the concept and entity layers a curated map of what the corpus is about, rather than a thin page per passing mention.
4. Log — every fetch, ingest, query, and lint operation is appended to log.md, so the wiki's history can be reconstructed from top to bottom.

Because raw sources are immutable and summaries link back to them, the wiki can be re-derived, audited, and extended without ever losing provenance.

Navigating the Knowledge

• By topic — browse the concepts index for the recurring techniques, patterns, and ideas the corpus teaches (AI governance, quality gates, platform engineering, retrieval, and more).
• By who — browse the entities index for the people, organizations, products, and places that recur.
• By source — browse the sources index for one summary per ingested reference.
• By question — filed Q&A sessions live in queries/.

Cross-links are standard, bundle-relative markdown links (for example, AI governance), so the graph stays portable across editors, static site generators, and agents.

Contributing

The wiki is maintained by agents and humans following AGENTS.md. Before adding content, read that manual: it defines the directory structure, OKF conformance rules, naming and frontmatter conventions, the ingest/query/lint workflows, and the quality bar. In short — never edit references/; summarize it in sources/, distill it into concepts/ and entities/, link bidirectionally, and log the operation.