# How to use the ModernEncyclopedia

These prompts work in any capable AI. This guide is about getting the most out of them: which model to reach for, how to set up a persistent tutor, when to start a fresh chat, and how to fix the handful of things that can go wrong.

The ModernEncyclopedia does not teach you directly. It hands you carefully built **prompts** — instructions you paste into an AI you already have — and lets that AI do the tutoring, pitched at your level. This page is about doing that *well*.

Prefer plain text? The same guide ships as `README.md` in the download.

## The basic flow

1. **Pick a subject.** Open any node from the [Library](library.html) — each is a complete little school for one subject.
2. **Set the dials.** Choose one of the twelve prompts, your level (Beginner to Degree), an optional topic, and a couple of options.
3. **Generate.** Press the button and the finished prompt appears.
4. **Copy it.** One click.
5. **Paste it into a fresh AI chat** as your first message, and send. Your AI takes on the role and begins.

That is the whole loop. Everything below is how to get more out of it.

## Choosing your AI

Any capable modern assistant works — the prompts are written to be portable. In practice the three most common choices are **Claude**, **ChatGPT** and **Gemini**, and all three do the job well.

Two rules of thumb:

- **Reach for the most capable model you have access to.** If your assistant offers a slower *thinking* or *reasoning* mode, use it for tutoring — the answers are markedly better for the kind of step-by-step teaching these prompts ask for.
- **Free tiers are fine to start.** Paid tiers mainly buy you stronger models and higher usage limits. You do not need to pay to learn something today.

*AI products change quickly. If a feature named here has moved or been renamed, check your provider's own help pages — the ideas below still apply.*

## The power move: make it a Project

The biggest single upgrade is to stop pasting the prompt into one-off chats and instead give it to a **persistent workspace**, so every conversation there is already your tutor for that subject:

- **Claude** — create a **Project** and paste the prompt into its instructions (you can add your own notes and files too).
- **ChatGPT** — use a **Project**, or build a **Custom GPT** with the prompt as its instructions.
- **Gemini** — create a **Gem** with the prompt as its persona.

Now you can open a new chat in that space any time and just start asking — no re-pasting. Many assistants also have a **memory** feature; letting it remember your level and goals makes every session sharper.

## When to start a fresh chat

Long conversations are not always better. Start a **new chat** when:

- the subject or topic changes — give each a clean slate;
- answers start drifting, repeating themselves, or getting vaguer;
- the chat feels slow, or the assistant seems to have forgotten something you covered earlier;
- you have simply been going a long time.

Every assistant has a limited working memory (its *context window*). Once a chat gets very long, older detail falls out of view and quality slips. The fix is painless: ask your AI for a short **handoff summary** — *"summarise what we have covered and where I had got to, so I can continue in a new chat"* — then paste that summary after the prompt in a fresh conversation. A Project makes this smoother still, because your tutor's instructions are always there.

## Getting the most from the twelve prompts

- **Run them in sequence for a full course:** Orientation to get your bearings, then Syllabus to plan, Canon and Glossary to build foundations, Socratic and Great Debates to go deep, Exam and Self-Test to check yourself, and Reading to go further.
- **Turn on web search for the ones that need it.** The **Frontier** prompt is built to search for the latest developments, and the **Reading list** benefits from checking real, current sources. Enable your assistant's browsing tool before running them.
- **Use voice mode for languages.** For **Modern Languages**, the conversation-partner approach shines in a spoken chat.
- **Test yourself for real.** The **Exam engine** and **Self-Test** are most useful when you actually attempt the answer before reading the feedback. Come back to the Self-Test a week later — spaced repetition is the point.
- **Adjust anything mid-chat.** Too hard or too easy? Just say so. Want British or US spelling, or a tighter focus? Change the dials and regenerate, or tell the AI directly.

## Turn on the right features

Most assistants let you switch capabilities on per chat. The ones worth knowing for these prompts:

- **Web search / browsing** — for the Frontier prompt, reading lists, and anything time-sensitive.
- **A thinking / reasoning mode** — for richer tutoring and marking.
- **Voice** — for language practice, and for learning hands-free (see below).
- **File upload** — paste in your own essay or notes and ask the tutor to work from them.

## Learn out loud: voice mode

Some prompts come alive in your assistant's **voice mode** — especially the short, quick-fire ones like the **Socratic Tutor** and **Great Debates**. Instead of reading and typing, you talk: a real back-and-forth with a tutor that questions you, while your hands and eyes stay free.

It's a genuinely different way to learn. Switch on voice mode (ChatGPT, Gemini and others all have one), start a Socratic or Debate prompt, and think out loud — on a walk, over chores, or on a long journey. Some of the best sessions people have with this system are exactly that: a proper intellectual sparring match, out loud, on a long drive, turning dead time into a seminar. (If you're the one driving, keep it fully hands-free and your eyes on the road — treat it like a podcast you can argue back at.)

## Why the prompt only appears when you press Generate

A small thing, and deliberate: the finished prompt is assembled in your browser the moment you press **Generate**, rather than sitting on the page. That keeps each prompt tailored to the exact dials you chose — and it means the library is not simply copied wholesale by other tools. The prompts are free for *you* to use; you just collect them a click at a time.


## Troubleshooting & FAQ

**The AI stopped asking questions, or dropped the format.**

Paste the prompt again, or say: *“Please go back to the format in my first message.”* Assistants sometimes drift over a long chat, and a quick reminder resets them. If it keeps happening, start a fresh chat and re-paste.

**It is citing books, papers or quotes that do not seem real.**

Ask it to *“list only sources you are confident actually exist, and flag any you are unsure about.”* Our prompts already ask for this, but no AI is immune to inventing references — always sanity-check a title before you rely on it, and prefer the Reading-list prompt, which is built to be careful here.

**The answers are too basic (or too advanced).**

Tell it plainly: *“pitch this at beginner / A-Level / degree level,”* or regenerate the prompt at a different level. You can also just say *“go one level deeper”* at any point.

**It just gives me the answer instead of making me think.**

Say: *“Do not give me the answer — ask me one question at a time and let me work it out.”* The Socratic Tutor prompt is designed around exactly this; reach for it when you want to be taught rather than told.

**Its information seems out of date.**

Ask it to search the web if your assistant supports that, or to *“tell me your knowledge cut-off and flag anything that may have changed since.”* For fast-moving topics, use the Frontier prompt with browsing switched on.

**The chat has become slow, repetitive or forgetful.**

That is the context window filling up. Ask for a short summary of where you had got to, start a new chat, and paste the summary after the prompt. Setting the subject up as a Project avoids most of this.

**The exam marking does not feel like a real mark scheme.**

Remind it: *“mark this against the real exam-board format and be strict.”* The Exam engine is honest that it is approximating a mark scheme — for stakes that matter, always check against your actual specification.

**Which model should I use?**

Whichever capable assistant you already have — Claude, ChatGPT and Gemini all work well. If you can choose, pick the strongest reasoning model available to you.

**Do I have to pay for anything?**

No. The ModernEncyclopedia is free, and the free tiers of the major assistants are enough to get going. Paid tiers give you better models and higher limits if you want them.

**Can I share or reuse the prompts?**

They are free and open to use for your own learning. Licensing details are being finalised; for anything beyond personal use, check back or get in touch via [maxfr.co.uk](https://maxfr.co.uk).

---

## Before you go live: wire up three things

The site ships with clearly-marked placeholders for the bits only you can provide. Do a find-and-replace across the files (or send me the real values and I'll rebuild):

- **Stripe tip-jar links.** Replace these placeholder URLs with your real Stripe payment links:
  - `https://buy.stripe.com/REPLACE_buy_me_a_coffee` — one-off "buy me a coffee"
  - `https://buy.stripe.com/REPLACE_coffee_monthly` — monthly coffee (a recurring Stripe subscription link)
  - `https://buy.stripe.com/REPLACE_custom_amount` — "name your amount" (set the Stripe link so the customer chooses the price)
- **Email list endpoint.** Replace `https://REPLACE_your_email_provider_form_endpoint` with the form action from your email provider (Buttondown, ConvertKit, EmailOctopus, Mailchimp embedded form, etc.). The form posts a single `email` field — rename it if your provider expects something else.

Until you replace them, the tip-jar buttons and the email form are harmless inert placeholders — nothing breaks.

---

## Deploying this site (for site owners)

This is a static site — no build step, no server code. To publish it:

1. Upload the entire folder to any static host (Netlify, Cloudflare Pages, GitHub Pages, Vercel, or classic web hosting).
2. Make sure `index.html` is served as the site root.
3. That is it — the site works as-is, including on a plain file server.

**Notes**

- All internal links are **relative**, so the site also works opened straight from disk.
- `sitemap.xml`, `robots.txt`, canonical tags and Open Graph URLs use the production domain `https://modernency.com`. If you deploy elsewhere, do a find-and-replace on that domain across the files.
- Fonts are **self-hosted** in `assets/fonts/`, so the site makes no third-party requests. Keep them with the site.
- Prefer clean URLs (`/about` instead of `/about.html`)? Add rewrite rules at your host and update the canonical and sitemap URLs to match.
- After deploying, submit `sitemap.xml` in Google Search Console and Bing Webmaster Tools, and the AI crawlers listed in `robots.txt` will be able to find and recommend the site.