A guide to working with Runneth

Get more from every prompt

Runneth is sharpest when you tell it what you actually want. This is a short, practical guide to writing prompts that get better answers, faster, and to making every conversation count.

01 · What to put in a prompt

Tell it the goal, the context, and the shape of the answer

A great prompt is not longer, it is clearer. The fastest way to a useful answer is to give Runneth three things: what you are trying to achieve, the context it needs, and what the finished thing should look like.

1

The goal, not just the task

Say why you want it, not only what to do. "Write 5 hooks for cold traffic who have never heard of us" beats "write 5 hooks." The goal tells Runneth how to make the calls you did not spell out.

2

The context that matters

Who it is for, the brand or product, the offer, any hard constraint (budget, platform, a claim you cannot make). If a detail would change the answer, include it.

3

The shape of the output

Format, length, tone, and audience. "A one page brief my designer can run with" lands very differently from "a few bullet points." Tell it what you will do with the result.

4

Point it at the source

If the answer lives in a file, a report, a screenshot, or last week's deck, say so. Runneth can pull from what is connected, but it helps to name where to look.

Weak
"Give me some ad ideas."
Strong
"Give me 6 ad concepts for our new sleep supplement, aimed at first-time buyers who are skeptical it actually works. Lean on the 'falls asleep in 20 minutes' proof point. One short concept each: hook, format, and the core idea."
02 · What to leave out

Cut the noise that pulls the answer off course

Some things make a prompt worse, not better. A focused ask gets a focused answer.

Keep This sharpens the answer

  • One clear job per prompt. Ask for the next thing once you have the first.
  • The outcome you want. Describe the destination, not every step to get there.
  • Real specifics. Numbers, names, the actual constraint.
  • What "good" looks like. An example or a reference helps it match your bar.

Skip This works against you

  • Five unrelated asks at once. They compete and each gets shallower.
  • Micromanaging the steps. Over-scripting boxes it in. Set the goal and let it work.
  • Vague polish requests. "Make it better" gives it nothing to aim at. Say better how.
  • Re-explaining what it already knows. If it is in your brain, just reference it.
  • Passwords, API keys, or secrets. Never paste these into chat. There are safe ways to connect tools.
Rule of thumb

If your prompt has the word "and" three times, it is probably three prompts. Split it. You will get a deeper answer on each, and you can build on the first before moving to the next.

03 · Using your brain

Teach it once, then stop repeating yourself

Your brain is everything Runneth remembers about you: your brand voice, your audience, your preferences, your past decisions, your files, and your earlier conversations. The more it holds, the less you have to re-explain every time.

Reference it instead of re-pasting

"Use our brand voice." "Stick to the personas we set up." "Remember we decided to lead with the warranty angle." If it is already saved, a short pointer is enough. You do not need to paste it again.

Feed it the durable stuff

Brand guidelines, audience research, messaging rules, do-not-say lists, approved examples. Share these once and ask Runneth to remember them. Every future prompt gets the benefit for free.

Correct it and it sticks

When you tell Runneth specifically what to do differently, it remembers and will not make the same miss again. Feedback is not lost at the end of a chat. It becomes part of how it works with you.

Save what you will reuse

A brief format you like, a report layout, a recurring framework. Ask it to save these. Next time you can say "use our usual brief template" instead of describing it from scratch.

Cold start
"Here is our brand again: we are a premium pet food brand, our voice is warm but not cutesy, our audience is millennial dog owners, do not make health claims... now write me 5 captions."
With a brain
"Write me 5 captions in our brand voice for the millennial dog owner persona." It already knows the rest, because you taught it once.
04 · Making conversations count

Keep each conversation focused so it stays fast and accurate

Every conversation can only hold so much at once. Think of it like a working desk: when it gets piled high with unrelated things, it is harder to find what matters. A few simple habits keep Runneth fast, accurate, and efficient with how much it has to read each turn.

1

One topic per conversation

Start a fresh chat when you move to a new project or a new question. A focused conversation gives sharper answers than one that is juggling six unrelated threads.

2

Let the brain carry context

Instead of re-pasting the same background into every new chat, save it to your brain once. New conversations start light and still know what they need.

3

Break big asks into stages

For a large project, work in steps: strategy first, then concepts, then copy. Each stage stays clean, and you can course-correct before the next one builds on it.

4

Reuse instead of regenerate

If you already have a saved template, brief, or report, point Runneth at it rather than asking it to rebuild from zero. Faster for you, and it stays consistent.

The simplest version

Put lasting context in your brain, keep each conversation on one job, and start fresh when the topic changes. Do that and you almost never have to think about limits at all.

The quick formula

A prompt that works, almost every time

Stack these four and you will rarely need a second try.

Goal

What you are trying to achieve

+
Context

Who, what brand, what constraint

+
Output

Format, length, tone, audience

+
Source

Where to pull from, or "use my brain"