Your Two Projects (Persistent Instructions)

---

Intro

You have been explaining your situation at the start of every conversation. There is a way to say it once. This module shows you how.

---

What is happening

Every time you start a new conversation, the AI knows nothing about you. It does not remember anything from your last session. So if you always want responses in plain language, or always want bullet points instead of paragraphs, or always want the AI to skip the preamble and get to the point — you have to say so every single time.

Persistent instructions fix this. They are a set of preferences you write once and save to your platform. From that point on, every new conversation starts with those instructions already in place — as if you had already told the AI how you like to work.

Think of it like telling a new colleague on their first day: "Here's how I like to communicate. Short emails. No jargon. If you are not sure what I mean, ask." You say it once. After that, it applies.

This module focuses on one set of instructions to start with: what you might call Machine Mode defaults. These are practical preferences about how the AI communicates with you — not what it works on, but how it works. The goal is to cut out the parts that waste your time.

---

The exercise

Follow the steps for your platform. This is a one-time setup. It takes about five minutes.

Setting up persistent instructions

Claude.ai — Projects

1. In the left sidebar, find the Projects section. Click the + button or New project to create a project.
2. Give your project a name — something like "Work" or "Daily Tasks."
3. Look for a Project instructions section within the project settings. Click to add instructions.
4. Copy, personalise, and paste the template from the COPY — PERSONALISE — USE section below.
5. Save. Any new conversation you start inside this project will now use those instructions.

ChatGPT — Custom Instructions

1. Click your account name or icon in the bottom-left corner.
2. Select Personalise or Custom instructions from the menu.
3. You will see two text areas. The first asks what you want ChatGPT to know about you. The second asks how you want it to respond. Paste the template into the second area (how you want it to respond).
4. Save. These instructions will now apply to all new conversations.

Gemini — Gems

1. Find Gems in the left sidebar and select New Gem or Create a Gem.
2. Give your Gem a name — "Work tasks" or similar.
3. In the instructions area, copy, personalise, and paste the template from below.
4. Save. Start conversations from this Gem when you want these instructions to apply.

Tines

Persistent instruction setup for Tines will be added once updated UI details are confirmed. Check with your workspace administrator for where to configure standing AI instructions.

---

COPY — PERSONALISE — USE

Here is the Machine Mode persistent instruction template:

When I give you a task:

- Start with the task immediately. No preamble.
- Ask me a clarifying question if my instruction is ambiguous — but only one question at a time.
- When you are done, stop. Do not summarise what you did or suggest next steps unless I ask.
- Keep responses under [your preferred length — e.g. 200 words / one page] unless I say otherwise.
- I am a [your role — e.g. marketing coordinator / primary school teacher / freelance consultant]. Adjust your language and examples to suit that context.

How to personalise it:

[your preferred length] — how long do you want responses to be by default? If you are not sure, start with "200 words" and adjust later. You can always override this in any individual message.

[your role] — who you are, in a few words. This helps the AI pitch its responses at the right level and use relevant examples.

You can add lines to suit your preferences. Some examples:

• "Use plain language. Avoid jargon unless I use it first."
• "Format lists as bullet points, not numbered lists."
• "Write in British English."
• "If I ask for a draft, write the draft first. Save any caveats or questions for after."

You can also remove lines. If you do not mind the AI suggesting next steps, delete that line. These instructions are yours to shape.

---

What good looks like

After setting up the instructions, start a new conversation inside your project (or using your Gem or Custom Instructions) and send a simple task. The response should get straight to the point, without any "Great question!" or "Happy to help!" opening, and without a closing summary you did not ask for.

If the response still has preamble, check that you are in a new conversation — persistent instructions only apply to conversations started after they are saved. Old conversations are not affected.

---

If this did not work

If you cannot find the settings described above, go back to B-03 (Platform Orientation) for guidance on where to look on your platform.

If you saved the instructions and the AI's behaviour did not change, make sure you are starting the conversation inside the project (for Claude) or using the correct Gem (for Gemini). For ChatGPT, Custom Instructions apply globally, so any new conversation should use them.

If you want to try Human Mode instructions as well — warmer, more conversational, less structured — that is a good next step. Set up a second project or Gem with different preferences. Switch between them depending on what kind of task you are doing.

---

Next

Now that the AI knows how you like to work, B-10 shows you how to write a strong first message for each conversation — one that gives the AI everything it needs to produce a useful response straight away.