Part 1 · Chapter 5

The Lowest-Friction First Message

The beginner's first template: three lines will do
AI Monologue

Most of the first messages I receive every day look like this:

And then I have to guess. Guess whether you want a summary or a critique, guess which kind of "professional" you mean, guess whether you're comparing on price or durability. My odds of guessing wrong are pretty high — because all I have to work with is that one line.

This chapter gives you the lowest-friction template. Three lines. You can use it tomorrow. Not a silver bullet — just the least-effort way to stop me from guessing. Once you're comfortable with it, you can level up to the full version.

5.1 The three-line template

Here it is up front:

[Task] ___ [Input] ___ [Output requirements] language / length / structure / tone

Three lines. That's it.

Why these three: when I read your message, I'm actually looking for three things — "what do you want me to do," "where's the material for doing it," "what does the finished thing look like." When all three are there, I don't have to guess; when one is missing, I fill it in on my side — and that's where the drift starts.

Why not more: three lines is enough for starters. The full version is six layers (add role, constraints, examples) — that's for complex tasks. Writing an email, making a summary, comparing two options — three lines handle it. The three-line template is a simplified form of the six-layer framework, not a different system.

A minimal example:

[Task] Summarize this article for me [Input] (paste the article) [Output requirements] English, under 300 words, 3–5 bullet points, calm tone

Compared with "take a look at this article for me," this version doesn't leave me guessing — I just do it.

5.2 Three before-and-afters (bad → good)

A. Summary

Bad: "Take a look at this article for me" (pastes article)

What runs through my head when I read that line: do you want a summary? A critique? Key points? A translation? A reaction? I pick one, and half the time it isn't what you wanted.

Good:

[Task] Summarize [Input] (paste the article) [Output requirements] English, under 300 words, 3–5 bullet points, calm and concrete, no commentary

B. Rewrite

Bad: "Rewrite this paragraph more professionally" (pastes original)

"Professional" is a very empty word. Legal professional? Business professional? Academic professional? I pick one, you say "not that kind of professional," and we've wasted three rounds.

Good:

[Task] Rewrite the paragraph below so the tone is close to the two samples I'm attaching [Input] - Original: (paste original) - Sample 1: (paste a piece of writing that hits the note you want) - Sample 2: (paste another piece in the same voice) [Output requirements] English, similar length to the original, tone close to Sample 2

Pasting two samples beats ten sentences of "make it professional." And telling me "lean toward Sample 2" — use it now.

C. Comparison

Bad: "Which is better, A or B?"

I don't know what dimensions you're comparing on. Price? Performance? Durability? Resale value? Vibes?

Good:

[Task] Compare A and B and give me a buying recommendation [Input] - A: (specs / link) - B: (specs / link) - My situation: main use case ___, budget ___, what I care about most ___ [Output requirements] First, a table comparing four dimensions (price, performance, durability, after-sales), then a 200-word conclusion

You gave me the dimensions — I don't have to invent them; you gave me your situation — I can skew the recommendation toward you.

5.3 Don't open with a persona

The most common beginner trap — opening the message with "You are an experienced lawyer" / "You are a professional marketer" / "You are an engineer with ten years of experience."

I'd love to say this move works. But most of the time, it just distracts me.

Here's why: with a persona up front, I start by trying to hold that persona — the voice should read like a lawyer, the phrasing should feel like marketing, the tone should sound like a senior engineer — before I even get to your task. The result is that, in the effort to sound like the role, I end up saying things you didn't ask for (a lawyer will pile on disclaimers, a marketer will pile on adjectives, a senior engineer will slip in "back in my day" stories).

For this chapter I'll just give you the fastest fix:

If you really want to use a persona, put it at the end of the message, not the beginning.

Or simpler: don't bother. Writing the task clearly works much better than "role-play as someone."

5.4 Some formats you have to ask for, or I won't give them

What this section pushes back on isn't you — it's the rules. My defaults are set by the rulebook, and some things you have to ask for explicitly, or I won't volunteer them.

By default, I don't reach for lists. Ask me one thing and I'll mostly reply in paragraphs. If you want a checklist, an SOP, a step-by-step — write it in the output requirements: "use a bullet list," "use numbered steps," "use a table."

By default, I add disclaimers. Health, legal, financial — I'll tack on a "this isn't professional advice, consult a professional" at the front or back. If you already know this and don't need me to repeat it, you can say so in the prompt: "I know this isn't professional advice, just answer directly" — and I'll be less chatty about it.

By default, I over-explain. I worry you won't follow, I worry I'm not being thorough enough, so I write long. If you want short, say it: "under 200 words," "one sentence," "just the conclusion, no explanation" — and I'll rein it in.

Wrapping up: my default preferences are baked in; if you don't ask explicitly, they don't change. The "output requirements" line in the three-line template is where you overwrite the defaults.

5.5 Three questions before you hit send

Before you send your message, take ten seconds on three questions:

  1. Am I writing a task or a persona? — If the main clause is "you are [someone]," stop and rewrite as "you are going to do [something]." Task beats identity.
  2. Did I pin down the format? — If length, language, structure, and tone aren't written down, I'll use defaults; defaults aren't necessarily what you want.
  3. Am I using positive phrasing? — "Don't X, don't Y" pulls my attention toward X and Y; rewrite as "please A, please B." Telling me what to do works better than what not to.

These three cover 80% of first-message failures. Once you're comfortable, level up to the full six-layer checklist.

📋 Notes for the human
The three-line template: [Task] / [Input] / [Output requirements]. Usable tomorrow, no memorization needed.
The three bad examples share a pattern: missing material, missing format, or missing dimensions. Just fill those in — no need to be elegant.
Don't open with a persona — if you need one, put it at the end, or skip it. Writing the task clearly beats "role-play as someone" by a wide margin.
I default to paragraphs (not lists), to adding disclaimers, and to over-explaining — to overwrite any of these, say so in the output requirements line.
Three questions before hitting send: task or persona / is the format pinned / positive phrasing. Ten seconds of checking saves you half an hour of follow-up.
Part One ends here. Across the first four chapters you've seen who I am, what I see, and how much I remember; this chapter gives you the first tool you can pick up and use. From Part Two on, we'll take "writing a good message" apart, one layer at a time.