Build, Document, Share — The Minbook Manifesto

“With AI tools, anyone can build.” “The era of the one-person unicorn.” The social timeline is full of these claims. Honestly, FOMO hit. As someone from a non-tech background, could I really build and operate a product with one click? I started this journey carrying that doubt and that expectation in equal weight.

What I found, once I jumped in, was that “one click” is the beginning, not the end. Claude and Cursor will write code for you, but how that code actually behaves in production, how to wire up payments, why security breaks — past the tutorials, a long line of messy problems is waiting. Minbook is the personal execution notebook from inside that mess.

Why Another Blog?

There are already plenty of good AI blogs. But once you try to ship your own SaaS, you find a gap between two kinds of material. One side: marketing-heavy content full of concept explainers and news summaries. The other: framework docs and academic papers. In the middle — the question “how do I apply this idea to the system I’m operating right now” — is surprisingly underserved.

Minbook is written to fill that gap. Every post tries to answer one of three questions:

• Where will this tool or framework break when I move it into my own system?
• How should this market trend reshape my decisions?
• Does this structural pattern reproduce in other markets?

All three require going one step past the headline. Minbook fills that step with first-person notes from actually walking into the wall.

Build, Document, Share — The Editorial Philosophy

These three words organize the editorial philosophy. Without making each of them concrete, content drifts toward abstraction fast.

Build

Not just observing code — building products that actually run. WICHI, the GEO SaaS, started as a hackathon rejection and reached production, with the 9-Bucket framework and multi-engine architecture stacked from the ground up. Make Me Unicorn (MMU), an open-source CLI, standardizes 534+ SaaS launch items. Self-Tuning Loop is a self-improving system that recovers learning signal between AI drafts and final outputs. This site itself (minbook.dev) is content infrastructure built from scratch — Astro · Tailwind · MDX · Vercel — with auto-publishing, OG image generation, and bilingual routing all running as code.

Document

The slog hidden behind “one click” gets recorded. The lab is generous; ops is not — auth, payments, infra, monitoring, security — and the goal is to put those operational stages into a single post as primary-source material. The decisions tutorials skip are exactly what this notebook captures. Average post length is 8,000–12,000 Korean characters (approximately 5,000 English words), with infographics, diagrams, code snippets, and tables. Even at that length, the structure is built so a reader can re-open the piece later as a decision reference, with clear H2/H3 hierarchy.

Share

I want the trial-and-error of a non-tech-background builder to become a marker for someone else. If the time I spent lost saves you a shortcut, that’s enough. Every post is published in both Korean and English, and AI search bots (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and others) are explicitly allowed to index everything. The work is designed to land for English-speaking readers in AI search, not just Korean web search.

What’s Covered — Three Categories

Minbook’s content lives in three categories. The starting rule: clearly separate “what I built” from “what others built.”

• Agents & Architecture — Inside the design of frontier agent tools: frameworks, orchestration patterns, memory management, Claude Code. Distilled into patterns transferable to your own system.
• Strategy & Economics — Data-driven analysis of the $660B AI market, capital flows, and the structural specifics of Korea and Asia. Built on primary sources (WIPO, OECD, IR filings) — looking past headline ARR into GPU margins, license fees, and undisclosed revenue mix.
• Build Logs & SaaS — Live data and PMF logs from products built and operated firsthand (WICHI · MMU · Self-Tuning Loop · Minbook itself). Closer to decision notes than code.

The rule is simple: what I built goes to build-logs, what others built to analysis, agent theory and architecture to agents. This split keeps the objectivity of strategy pieces and the inside view of build logs both alive.

Going Deep with Series

More than 80% of Minbook posts are bundled into series — three or more pieces digging into a single subject. A series helps decisions more than a one-off post does. Eleven series are running: Claude Code Anatomy, Ralph Loop, Self-Tuning Loop, Agent Orchestration, AI Monetization, AI Market Anatomy, GEO Paper Review, GEO Foundations, WICHI Build Log, MMU Build Log, Korea SaaS Build Trap.

Each series has its own index at /en/series, and within a series, prev/next navigation is auto-wired into every post. Each series page also opens with “why this series exists / who it’s for / how to read it” — useful for finding an entry point.

Editorial Operating Principles — Trust Starts with Transparency

Transparency is where trust starts. The full writing/verification/correction policy lives at Editorial Policy; the core points:

• Primary sources first — official docs, IR filings, academic papers, and GitHub repos take precedence over secondhand articles. Numerical, performance, and pricing claims always carry a clear timestamped source.
• AI-augmented disclosure — every post is finished through dense collaboration with AI. AI handles research aggregation, KO↔EN translation, and draft rewriting; the direction of questions, contextual insight, and final fact-checking remain with the operator (M). AI-fabricated sources (hallucinations) never get published unverified.
• Conflict separation — all content represents personal views only and does not represent any current employer, client, or partner. Topics with direct stake are not covered.
• Revenue model — Google AdSense auto-ads cover domain and infrastructure costs. Ads are visually separated from content. No sponsored content. Affiliate links, when used, are marked in the body.
• Correction policy — factual errors get corrected immediately with the “last updated” timestamp refreshed; logic-breaking errors get a “Correction Notice” at the top of the piece. Reader corrections are accepted at contact@minbook.dev.

Who This Is For

• Solo builders shipping their own SaaS or OSS
• PMs and architects evaluating internal AI agent rollouts
• Analysts and researchers treating AI tool architecture as study material
• Strategy leads looking for the structural opportunity in Korean and Asian AI markets
• Anyone wanting to answer “is AI a bubble?” with data

The fit is sharpest when you need a decision frame for “why this choice” more than a tutorial line to copy. For operators and investors of Korean SaaS and AI companies, the coverage of Korea’s structural ceiling and ceiling-break together is harder to find in other English-language outlets.

How to Read

If this is your first visit:

1. Skim the first five posts in the category closest to your interest
2. Pick the one that pulls hardest, then follow the series banner at the bottom of the post into deeper material in the same series
3. Browse the series index for active and completed series side by side
4. Subscribe to RSS (/en/rss.xml) for new-post notifications

If you only want a specific project case, the WICHI project page and the MMU project page consolidate the build logs in chronological order.

Closing

Most tutorials end at “deployed.” Real development starts there — that’s the lesson learned by walking into walls. Minbook is the place that goes deep into “after the deploy.”

If you’re working through similar problems, I hope something in these notes becomes a hint. Feedback, corrections, and collaboration proposals are always welcome at contact@minbook.dev.