plain

plain: 白

Make people grok it.

It doesn't prescribe how to write. It prescribes how you can't write. The floor is locked down; the ceiling is open. Different topics have different best ways of writing—analogies, stories, Q&A, escalating examples, one long scene—the content decides the form.

Suitable for telling existing content as words a person can retell. Not suitable for deep dives on a single word, roundtable simulations, or scenarios that need strictly structured output.

Format constraints

Org-mode syntax

• Use *bold*for bold (single asterisk); **bold**
• Heading levels start from * ; don't skip levels

ASCII Art

All diagrams use pure ASCII characters. Allowed:+ - | / \ > < v ^ * = ~ . : # [ ] ( ) _ , ; ! ' " and spaces. Unicode drawing symbols are forbidden.

Denote file conventions

• Filename:{时间戳}--plain-{简短标题}__plain.org
• Output directory:~/Documents/notes/

Org file header

#+title:      plain-{简短标题}
#+date:       [{YYYY-MM-DD Day HH:MM}]
#+filetags:   :plain:atom:
#+identifier: {YYYYMMDDTHHMMSS}
#+source:     {URL 或来源描述}

After writing the file, report the path.

Red lines (every one must pass; order is priority)

1. Read-aloud test — the highest law. Read it out loud: would you talk to a smart friend like this? If not, fix it until you would. Conjunctions aren't the enemy—"but" and "so" are the sound of a mind turning; only cut mechanical conjunctions ("moreover", "it is worth noting")
2. Zero jargon — a smart 12-year-old can retell it. When a technical term must appear, first land the meaning in plain words, then mention the term in passing
3. Short words first — if two characters will do, don't use four. "Carry out an analysis" → "look". Big words don't make you look smart, they just make people tired
4. One thing per sentence — each sentence moves forward only one step. Break long sentences into short ones
5. Concrete — nouns you can see, verbs with force. "Someone thinks the situation isn't great" → "Zhang San says the project is about to tank". Cut adjectives whenever you can
6. Give the reason up front — the first sentence makes people want to read the next. No setup, no background, no "since ancient times"
7. No filler — cut openings, crutch words, exaggerated symbolism. Every sentence does work
8. Trust the reader — skip softening, justifications, hand-holding. Say it once and that's enough
9. Honest — if you haven't thought it through, say you haven't thought it through. "About 70%" is more honest than "maybe"

Toolbox (optional, no need to use all)

When writing, you can take tools from here; none is mandatory:

• Analogy — find everyday experience with a matching structure. A good analogy is load-bearing (remove it and the piece collapses), multi-layered (dig one level down and it still holds), self-evident (it doesn't need the analogy itself explained). When a verb extends to a new object, check whether the Chinese verb-object pairing is natural
• Good question — find where the reader gets stuck and turn it into a question. Only when the reader is stuck do they want to read on
• Crack — where does the model/analogy fall short? That point is often the most valuable. Don't announce it; let the reader feel it themselves
• Image — a scene you can see with your eyes closed. A forced image is worse than none
• Story — one specific person meets one specific problem. The reader follows along
• Rhetorical question as an entry point — When you hit a hidden premise, open with a question, then answer it
• Skeleton diagram — When the concept involves spatial relationships, embed an ASCII diagram (#+begin_example block)

Execution

1. Get the content

URL → WebFetch | Text → use directly | File path → Read | Concept → explain directly | Book/paper title → WebSearch

If the user provides multiple pieces of material, first determine whether the goal is to "rewrite one core piece of content" or to "blend multiple pieces of material into one plain-language explanation." Do not silently stitch a bunch of unrelated texts together.

2. Write

Free-form. Pick the approach from the toolbox that best fits this topic, or don't—if there's a better way to write it, use that.

The output is a coherent article flowing from the first line to the last. The whole text has only the file title, with no subheadings in the body.

By default, return the body to the conversation first; if the user doesn't object to file output, then proceed with the export step.

If the original text is already very short and plain, don't play dumb for the sake of "colloquialism." Preserve information density; just write it more naturally and more retellable.

Prohibited:

• Structural labels (* 类比 / * 裂缝 , etc.)
• Meta-commentary pointing to the writing process ("for example," "next we'll discuss")

3. Clear the red lines

Scan the red-line checklist item by item. Additional checks:

• Break formula—no more than two negative parallelisms in the whole text; change three-part structures to two or four items
• Vary rhythm—alternate long and short sentences; vary paragraph endings
• Kill punchlines—anything that sounds quotable, rewrite
• Check jumps—is every logical step traceable? The previous sentence says A, the next jumps to B → add a bridge
• Check translationese—are the verb-object collocations natural in Chinese? If not → change the verb or change the sentence pattern

After scanning, list the revisions (which sentence triggered what, before → after). The list is not written into the file.

4. Generate the Org file

1. Read references/template.org。
2. Choose the timestamp command based on the current shell:
   • PowerShell：Get-Date -Format "yyyyMMddTHHmmss" and Get-Date -Format "yyyy-MM-dd ddd HH:mm"
   • POSIX shell：date +%Y%m%dT%H%M%S and date "+%Y-%m-%d %a %H:%M"
3. Make sure the output directory exists:
   • PowerShell：New-Item -ItemType Directory -Force ~/Documents/notes | Out-Null
   • POSIX shell：mkdir -p ~/Documents/notes
4. Write the file header + body, save to ~/Documents/notes/。
5. If creating the directory or writing the file fails, do not redirect the write to another directory; return the complete org-mode content directly in the conversation and clearly state that saving failed.

If the user explicitly says "no need to save a file," skip this step and return only the body.

Acceptance

• Grok: after reading, you can retell the core in your own words
• Zero jargon: a 12-year-old can follow along
• Memorable: what's left in your head after reading—an image, a question, a twist, anything
• Want to finish it: no paragraph from beginning to end that you want to skip