Prompt Engineering Academy
Three free graded courses that teach the skill this whole site runs on: getting exactly what you want out of AI. Slide lessons, quizzes with fresh questions every attempt, a final test per course, and a capstone you can actually use.
Slides read + best quiz scores across all three courses, stored in this browser. The grading scale:
How AI actually reads your words, and how to ask so you get what you meant.
Roles, examples, structure, reasoning, voice, the five tools that turn decent answers into work you can ship.
Stop retyping prompts and start owning a system: a profile, specialist agents, and explicit rules for who wins.
Quick explainers
Not ready for a course? The short version of everything this site does:
What is a portable AI identity?
A structured set of instructions that captures how you think, write and want to be answered, owned by you as plain text. Instead of re-teaching every new AI who you are, you paste your profile once and it behaves like it already knows you. Switch platforms next year? The profile comes along.
Why questionnaires alone aren't enough
People describe their style imperfectly. That's why the profile leans on writing samples: we measure sentence length, contractions, hedging, punctuation habits and recurring phrases from text you actually wrote. Stated preferences steer the answers you want; observed behavior steers the voice.
Compact vs Standard vs Full
Every profile generates in three sizes. Compact (~1,000-1,500 characters) fits the small custom-instruction boxes. Standard is the everyday version for project fields. Full carries the entire style model. Use it where instructions can be long, like Claude Projects. Check any prompt's size →
Writing AS you vs answering TO you
Your profile makes the distinction explicit. Drafting an email you'll send = the AI imitates your voice, greeting to sign-off. Explaining something to you = it uses your preferred format and depth without pretending to be you. Both are covered by different sections of the same prompt.
What a professional agent actually is
Not a costume, a structured definition: role, workflow, frameworks, required inputs, output format, guardrails and disclaimers. Our agents refuse to fabricate data, state their assumptions, and say when a licensed human should take over. The prompt is generated from that structure. Browse them →
Profile + Agent: who controls what
The agent controls WHAT the AI can do and how it approaches the work. Your profile controls WHO it's talking to and HOW to communicate. An explicit precedence order keeps style from eroding substance: required disclosures and analytical steps survive even a “keep it short” preference.
Installing your profile anywhere
Persistent-instruction platforms (ChatGPT, Claude, Gemini…) take it once in settings or a project. Everything else works with paste-at-start: your profile as the first message plus “use this as persistent context”. Platform-by-platform walk-throughs: Integration guides →
The privacy model
Private by default is structural here, not a setting: answers and writing samples live only in your account, publishing is a separate reviewed step, sharing an agent config never includes your style without its own consent box, and export/delete are one click each. Full policy →
Glossary in 60 seconds
System prompt: standing instructions an AI reads before your messages. Custom instructions: a platform's box for exactly that. Context window: how much text a model can consider at once. Token: roughly ¾ of a word. Memory: platform-managed notes about you, useful, but your profile is the version YOU control.