
What happened
Given a request for a four-circle ベン図 in mermaid, the AI said composite Venn diagrams could not be drawn and offered another chart type; a forced image-fed attempt produced an unreadable result, so the author hand-drew it.
Why it matters
The author notes mermaid's automatic layout shifts each time with four circles and text, so output is not stable — and concludes that the real AI-era skill is 言語化, turning an expected result into a precise prompt.
Summaries like this, every morning.
The starting point is the ikigai four-circle diagram, which the author assumed would be quick to reproduce in mermaid now that AI can write code. That assumption failed twice: the AI first declined the diagram type outright, and when pushed with an image and instructions to use a venn diagram, it produced something shaped like a diagram but with unusable text. Beyond the model's behavior, the author points to mermaid itself — the automatic layout recalculates each run, so with four circles and labels the output drifts and cannot be controlled.
The author's takeaway is not a tool review but a skill argument. Getting a usable result required specifying what could be omitted, which characters were allowed, and where text should be trimmed — in other words, precise 言語化 of the expected output. From that, the piece proposes a format: a problem statement with sample answers that separate the parts allowed to vary from the relationships that must not change, such as the rule that what you love and what you are paid for together constitute a vocation.
The author frames this as a playful exercise rather than a product critique, inviting readers to post their own prompts in the comments, and ties it to a daily workflow of generating code — control over that output, the author suggests, is what the language-training question is meant to build.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Uravation published a free 32-page PDF plus Excel file with 50 ChatGPT prompts, five each for 10 departments…

Writing in Skeptic, Erez Bataat argued that a Pain Axis found in open-source models, and the distress they sho…

A developer building an AI-driven workflow with Claude Code and Codex — where a Jira ticket triggers autonomou…

A Zenn explainer defines Perplexity as a metric for how confidently a language model predicts the next word, w…

Japan AI Development Center announced a Claude Code seminar for December 9, 2026, 14:00–16:00, at the Higashih…

株式会社Ikigai said on 2026年10月30日 it will start a service covering ChatGPT ads, SEO and LLMO together, promising…
