Nine AI skills to get ahead of most people

Executive overview

Most people use AI for simple tasks and accept mediocre outputs. The top 1% treat AI as a creative operating system — one they program, iterate against, and personalise deeply.

Nine skills separate casual users from power users. Each builds on the last, moving from basic prompting to a fully organised, self-improving AI workflow.

The real edge is not access to AI — it's knowing how to fight with it, curate its outputs, and teach it who you are.

Prompt engineering

• Every prompt needs four elements: role, context, command, and format.
• Define the role to tell AI which body of knowledge to draw from.
• More context about your situation produces more relevant output.
• Specific commands produce specific results — vague requests produce vague answers.
• Providing a world-class example for AI to pattern-match against is a shortcut to better output.

Taste curation

• Taste is knowing what great looks, sounds, and feels like — essential for selecting the best AI output from many options.
• Build a taste library: collect examples of world-class work in your field (social media, YouTube, GitHub, etc.).
• Develop precise language: the exact words you use shape AI output — "leader" and "boss" produce different results.
• Document your taste as universal rules: "write in ninth-grade English", "use similes over examples", "no em dashes".
• Written, repeatable rules lock in taste so AI can consistently match it.

Master prompt

• A master prompt is a document covering your role, context, and preferences — uploaded at the start of every session.
• Ask AI to act as an interviewer and generate the questions needed to build your master prompt, then answer them by voice-to-text.
• Save the master prompt as a PDF so it can be uploaded to any AI platform, including ones not yet built.
• This future-proofs your setup and eliminates starting from scratch on new platforms.

Output iteration

• Most people accept the first output. The top 1% iterate until it's right.
• Upload your master prompt first to prime the session.
• Give specific feedback — not "make it punchier" but "open with a reframe that addresses X".
• Use the canvas feature in ChatGPT to lock in a version and tweak it manually, then use that version as a template for future outputs.

System prompts

• A system prompt tells AI how to behave, not who you are — it's the repeatable instructions behind a workflow.
• Generate a system prompt by giving AI your best output and asking it to write the prompt that would have produced it.
• Save system prompts as PDFs for portability across platforms.
• In ChatGPT, convert a system prompt into a custom GPT to create a reusable, shareable tool for your team.

Using AI as a critic

• AI defaults to agreement. Deliberately instruct it to push back.
• Set the role as "devil's advocate" to stress-test assumptions and surface risks.
• Ask it to break down its criticism through first principles — this rebuilds answers from base truth.
• When AI surfaces a useful insight, capture it and update your master prompt so the principle persists.

Context compression

• Too much context overwhelms AI and degrades output quality.
• Paste all raw material, then prompt: "Summarise this with bullets for key facts, data, and stories — reduce to 10% of original size."
• Ask what was removed, in case critical context was stripped.
• Use the compressed output as the sole context in a new session window.

Knowledge-base gardening

• A messy AI workspace produces messy outputs — organise by project, not by session.
• Create a named project folder for each initiative and upload the master prompt and compressed context into it.
• Store system prompts in organised folders by department as PDFs for reuse.
• Keeping outputs as PDFs makes it easy to switch AI platforms without losing your setup.

Personalised learning

• Use AI to generate research papers on any topic, in conversational language, at a chosen grade level.
• Specify time constraints: "teach me X, I have seven minutes."
• Use the read-aloud feature to absorb content passively — at the gym, walking, commuting.
• AI is not a chatbot. It is a co-creation tool for goals, blind spots, difficult conversations, and continuous learning.