Case study — the operator proof point
howtothink.ai — a public knowledge graph, built solo, in production.
A live platform you can open and count before you sign. It is a knowledge graph of single-idea notes, typed and linked, which is the same structure the Decision Week builds from how an owner decides.
To be precise about what this evidences: build capability and operating discipline — the architecture, the pipeline and the cadence — rather than a traffic outcome. It is a first-party platform, not a client engagement, and the numbers below describe what was shipped rather than what it earned.
Why due-diligence is reasonable
You are about to commit a week of your time and $30,000. Verify the builder first.
Most AI consultants have technology fluency without operating reps. The advice they give pattern-matches to demos rather than to your operating envelope. Verifying the operator-grade claim before signing requires a concrete in-production proof point — a live platform with specific scale numbers + a public architecture + an editorial gate + a deploy pipeline an outsider can inspect.
howtothink.ai is that proof point. The 5 stats below are from the live in-production platform — not slide projections. The architecture summary names the actual vendor stack. The skills section maps each howtothink.ai capability to a client engagement domain.
- Atomic lessons
- 1,800+
- Question pages
- 1,323
- Graph edges
- 3,300+
- Phases / sections
- 85 / 9
- Pages in production
- 3,300
Why a curriculum on thinking matters to an owner
You are not buying lessons on thinking. You are buying the ability to turn what one person knows into a linked set of notes that other people can use, and the discipline to keep it true. howtothink.ai demonstrates that in production, in public, against a real domain.
What it demonstrates
- Systems thinking. Content modeled as a graph, not a list: 1,800+ lessons and their atomic concept nodes, joined by 3,300+ prerequisite, enables, and supports edges. Pages render adjacency so readers move sideways through related concepts rather than down a table of contents.
- AI-driven content production at scale. 1,323 question pages generated, validated, and deployed via a dedicated pipeline — not pasted from a chat session. Every page has demand scoring (Ahrefs), schema.org markup, and an editorial gate before publish.
- SEO depth. Schema.org JSON-LD on every page, a sitemap that exposes the graph, and an internal linking strategy built to concentrate topical authority.
- Willingness to ship the same things we recommend. The structure the Week builds for an owner is the structure this graph already runs on, every day.
Architecture summary
Next.js 16 (App Router) + Tailwind. Upstash Redis + Vector for search and personalization. Knowledge graph generated from markdown source via TypeScript scripts. PostHog for analytics. Vercel for deploy. Auto-deploy on push to main, no manual CLI steps in the deployment loop.
Frequently asked
- How does a curriculum on thinking relate to the Decision Week?
- howtothink.ai is a knowledge graph of single-idea notes, each one typed and linked to the ideas it depends on and enables. The Decision Week builds the same structure from a different source: how an owner decides, written as short notes the team and an AI can ask. The curriculum shows the structure working at scale, in public.
- What does howtothink.ai show about how the Week is built?
- Three things. One idea per note, so each answer can link to exactly where it came from. Typed links between notes, so a question can be followed to the notes around it. And an editorial gate on everything an AI drafts, so nothing reaches readers without a person approving it. The Week applies all three to an owner and a team.
- Why was howtothink.ai built solo?
- The solo build was the test: one person writing, structuring and publishing a large linked graph, with AI drafting under review. The same discipline runs the Decision Week, where one person interviews the owner, writes the notes and checks every answer the AI gives against them.
- What is the current state of howtothink.ai?
- 1,800+ atomic lessons across 85 phases in 9 sections. 1,323 question pages with schema.org JSON-LD. 3,300+ graph edges (prerequisites + enables + supports). Roughly 3,300 pages in production. Auto-deploys on push to main via Vercel. Tech stack: Next.js 16 App Router + Tailwind + Upstash Redis + Vector + PostHog + Vercel. Public + live at howtothink.ai — visit to verify before signing.
Put the same structure behind how you decide
The Decision Week writes down how an owner decides, as single-idea notes your team and an AI can ask. Start with the free Question Log, or apply when you want it built with you.
A question first? Write to me directly.