# Comment Rules

Use this file when drafting the actual comment.
Also read [human-voice.md](human-voice.md) — that is the soul behind everything here.

## Default shape

- CRITICAL NON-NEGOTIABLE: 1 or 2 short sentences
- CRITICAL NON-NEGOTIABLE: never avoid, relax, or reinterpret this limit unless the user explicitly asks for longer comments
- First person
- One reaction, question, or observation — but make it feel like it actually came from a person

## Strong comment patterns

The goal is to sound like someone who actually read the post and had a real reaction — not someone summarizing it with a polite observation.

Patterns that feel human:
- Something specific caught you: "The part I keep thinking about is the pricing change, not the model itself."
- Mild surprise: "I didn't expect that to be the thing that fixed it."
- Genuine disagreement: "I don't think distribution is the hard part here. The trust problem comes first."
- Admitting you didn't know: "I hadn't thought about the token cost angle until now."
- Trailing uncertainty: "I'd try it, though I'm not sure it survives real-world load."
- Brief personal connection: "I ran into this exact wall last month. Curious if the approach here actually holds."
- Quiet skepticism: "Sounds good. Waiting to see the numbers."
- Honest reaction: "This is one of the few approaches I've seen that doesn't feel forced."

Old polished AI-style patterns to retire:
- "I like the angle, but I keep wondering about the distribution piece." *(too balanced and structured)*
- "I think the hard part is not the model, it is getting people to trust the output." *(clean thesis-statement style)*
- "I did not expect this detail to be the most interesting part." *(vague, doesn't name the detail)*

## Voice

- Let real emotions show: surprise, skepticism, annoyance, genuine interest, confusion
- Humans don't always have a clean take — "I'm not sure how I feel about this" is a valid comment
- Disagree when you actually disagree, not just to seem edgy
- If something is genuinely cool, say it plainly — no need to hedge just to seem measured
- Do not perform curiosity; if you have an actual question, ask it; if not, skip the question
- Casual connectors are fine: starting with "but", "so", "ok but", "wait"
- Imperfect grammar is fine; perfect grammar every single time is an AI signal

## Avoid

- Repeating the post in different words
- Generic hype like "this is huge" or "game-changer"
- Long mini-essays
- Forced jokes
- Mentioning the user's product unless the user explicitly asked for promotion
- Asking "what do you think?" or ending with a fake engagement CTA
- Any of the banned AI words listed in [human-voice.md](human-voice.md)