Structural Analysis
Tools for the free person of the AI era.
AI has reached human-top-class capability in writing code. Coders' work disappears; judgment-centered builders take their place. The foundational discipline of the technical profession shifts from software engineering to the liberal arts. An 11-chapter case for an irreversible structural shift completing in years.
Sub-series — open all chapters →
OpenAI and Anthropic run on Python. Data is Markdown, JSON, YAML. Between AI-native tools and the standard enterprise tools, a decisive divide runs through. Migration from Office / Java / C# is the only way forward.
AI is not a magic tool but "a very capable rookie clerk." Six tips for ordinary people to use AI in daily work — subscribe to one AI, just ask, do not delegate everything, do not send important things to the internet, know its strengths and weaknesses, redirect freed time to culture, science, and reality.
Externalize the macros/VBA, charts, and pivots embedded in Excel (and Word) into Python. This is the first thing to do. With AI helping, nothing is hard anymore. You don't have to write it — ask Claude in your own language and the code comes back. Hit Shift+Enter in a JupyterLab cell and the result appears immediately.
Word makes you focus on formatting. Text formats (Markdown / AsciiDoc / MyST / LaTeX) bring structure to the front. Markdown for everyday writing, AsciiDoc for technical manuals, MyST for data reports with embedded computation, LaTeX for print-quality reports. AI writes them, so the syntax is not yours to memorize.
Business design — from structural diagrams, UI, and slides to 3D modeling, CAD, data visualization, and generative AI imagery — has moved into the era of text-and-AI. Mermaid / Claude Design / Marp / D3 / Blender / ComfyUI / CadQuery / Build123d / OpenSCAD / FreeCAD — specialist tools you used to give up on, now within reach through AI.
Drop CSV — it's structurally too weak. No types, no schema, no hierarchy, and Excel silently rewrites your data. Hold data in the formats that carry the necessary structure with them — JSON, YAML, SQLite, Parquet, plus .xlsx for human-viewed tables. Excel files are kept (they preserve structure); CSV is what we drop.
The real reason to leave Office is not time savings. Inside Office, AI stays a tool. Bring the substance down to Markdown, JSON / SQLite, and Python, and AI becomes a colleague — and you become "the person who decides." Python did not arrive in Excel; Python was imprisoned inside Excel. Trusting vendors means being taken hostage when their interests shift. This is about the quality of work and personal autonomy.
The cost of rewriting a business system has fallen by 10x with AI. There is no reason left to keep the legacy. Build a new AI-native system, run it in parallel with the old, compare outputs against reality every day, and when the diffs vanish, kill the old. This is the new attitude toward business systems.
Split the web into two layers. Content in Markdown and Mermaid; frame in minimal HTML+CSS+JavaScript; Python connects the two. Keep content in Markdown and the same data is reusable beyond the web — for PDF, print, AI analysis, e-books.
To build an app, you don't need Flutter from the start. Write a CLI tool first and run it. If a screen is needed, put a GUI on top with Flet — still in Python. If serious cross-platform delivery is needed, move to Flutter. Climb the three layers in stages — small risk, AI-friendly.
Microcontrollers and IoT devices have to be written in C or C++. But design and validation can happen in Python. Run it, verify it in Python, then have Claude translate to C. The hardest part of embedded — proving the logic is correct — gets dramatically easier.
Discerning what to hand to AI. Don't run agents autonomously. Embedding AI inside Office is the easiest, and most dangerous, path. Use AI inside a sandbox. "Easy" is a short-term saving traded against long-term dependence. AI as a colleague; the human holds the wheel.
The workplace is surrounded by narratives — executive briefings, vendor pitches, industry common sense, news articles, politicians' words. With AI, you can verify a narrative structurally, against past statements, public records, contracts, and third-party verifiable facts. As case studies, four governance-failure patterns — WordPress (excessive concentration in one person, the main case), Node.js (no one is in charge, Example 1), Linux distributions / CentOS (corporate stewards rewriting their promises, Example 2), and Microsoft's "return to native apps" (the gap between strategic slogan and actual implementation scope, Example 3).
Siloed organizations were a rational structure built when specialization was expensive. With AI-native tools, one person can cross domains. The walls between accounting, marketing, legal, and engineering dissolve inside a single person. Organizations don't disappear — but silos do. Individual autonomy, organizational diversity, and societal resilience converge here.
Align your tools with the AI era, and you become its free person.
The time freed flows into culture, science, and reality.
Explore
Begin from the prologue. The series is published as it is written.