📝 Getting started

Start using AI for real work in your first hour — a 2026 beginner's guide

You don't need to be technical to put AI to work. This walks you from “first chat” to “let AI touch real files and the web” in a single hour.

Getting started · Updated 24 Jul 2026 · ~7 min read

1. Pick the right AI

a comparison of popular AIs (Claude · ChatGPT · Gemini · Cursor) with short strength labels — Pick the AI that fits the job, then start

Writing/thinking/summarising → Claude or ChatGPT
Coding → Claude Code, Cursor
Cited research → Perplexity, Gemini

See what each tool supports at Supported AI.

2. Get good results (prompt basics)

State a role — “You are a Thai copy editor…”
Give context + examples — more input, sharper output
Specify the format — “answer in 5 bullets”

Don't want to write your own? Grab one from the Prompt catalog.

3. Give the AI skills

For work you repeat (code review, slides, writing), install a Skill once and the AI stays good at it — no re-explaining.

Browse the 278-skill catalog · new to Skills? Read Skills vs Prompts.

4. Let it touch real tools (MCP)

before/after — left: AI answering from memory (maybe wrong); right: AI via MCP reading the real file and answering correctly — MCP lets the AI work with real things, not guesses

Once comfortable, let the AI work with real things — local files, GitHub, the web — via MCP. Start with Filesystem + GitHub.

Setup in 10 MCPs worth trying.

5. Shortcut: use a Loadout

Don't want to assemble it piece by piece? A Loadout bundles Prompt + Skill + MCP + how-to for one job — download the .loadout.md and open it in your AI.

See the Loadout catalog.

Things to watch

Never put secrets/passwords in the chat or let the AI store them
Always check results before real use — AI still slips
Grant MCP least privilege; prefer official sources