Super Intelligence Academy

Demo prototype · Saved in this browser.Not synced — don’t enter sensitive information.

← Understand the Tools. Get Better Results.Lesson 7 of 7 · 15 min · Written lesson

Engineer useful context

Guide with constraints and examples

Loading your workspace…

The possibility

Write a request that shows the desired style while keeping facts, hard limits and unresolved details separate.

An example can show a rhythm, level of detail or useful structure more clearly than an adjective such as ‘friendly.’ But a sample also contains names, dates and promises that may not belong to the new task. Tell the assistant exactly what the example teaches and which source supplies the facts.

01
Name the hard requirements
Give requirements that can be checked: at most 70 words, include day and time, use only the current event facts. Keep preferences, such as warmth, separate. If a word limit conflicts with required content, ask for a shorter alternative or a clarification rather than silently dropping a requirement.
02
Annotate what to borrow
‘Use the sample’s welcoming opening and short sentences’ is more precise than ‘write like this.’ Add ‘do not copy its event facts or promises.’ The sample is an illustration of style; it is not evidence that the new event is free, equipped or open for reservations.
03
Supply one useful counterexample
A counterexample makes a subtle boundary visible. ‘All materials provided’ is wrong for an event that asks people to bring a pencil. Explain the failure, then ask for a version that obeys the current source. Avoid loading the request with so many bad examples that the actual task gets lost.
04
Inspect the output against separate checks
First check facts and missing details against the packet. Then check length and style. A warm tone cannot compensate for an invented price; source accuracy does not automatically make the invitation inviting. Ask a focused revision question and compare the changed sentence with its evidence.

This lesson uses fictional event details and authored examples. Its feedback is prepared, not generated from your request. Instructions help express what you want; they do not guarantee that an AI output will obey every constraint.