Choosing the Right Layer
A free lesson from Swipe AI · Prompt Engineering, level 2 · updated October 2026
By the Swipe team, founded by Colin · how we make lessons
Deciding where an instruction belongs comes down to one question: would you want this rule active for every future conversation, or does it only make sense for what you're asking right now? If the answer is 'every time,' it belongs in the persistent layer. If the answer is 'just this once,' it belongs in the user message.
Not every tool gives you a separate settings field for a persistent instruction. Even then, the two-layer thinking still helps: you can open a single prompt with your standing preferences, then follow with the specific task, so the two layers are still separated — just inside the same message instead of two different fields.
In practice, a good setup combines both layers on purpose. A support-reply assistant might have a persistent instruction covering tone, a standard closing line, and a length limit — while each user message supplies the actual customer complaint that changes every time. Neither layer tries to do the other's job.
Persistent instruction: "Always be empathetic and concise. End every reply with a note that help is available." User message #1: "Customer says their order arrived damaged." User message #2: "Customer says they were charged twice."
A team wants every reply from their AI assistant to always include a disclaimer and use a formal tone, no matter what the specific question is. Where should that rule live?
The simplest test for where an instruction belongs is asking whether it should apply to every future request or just the ____ one.
Even in tools without a separate persistent-instruction setting, it can still help to structure one prompt with standing preferences first and the specific task after.
Two Layers of Instruction
10 questions · pass with 60% — just like in the app.
What is a persistent instruction typically called in many AI tools?
A persistent instruction only applies to the single message it was written in.
The specific, changing request you type for one task is called the ____ message.
Which of the following is a standing rule well-suited to a persistent instruction?
A one-time task detail, like the specific document to summarize today, belongs in the user message.
A rule stuck in the persistent layer that only made sense for one past task will keep applying to ____ future requests too.
Even in a tool without a separate persistent-instruction setting, what can still help?
Combining a persistent instruction for tone with per-message task details is a good setup for a reusable assistant.
The simplest test for where a rule belongs is whether it should apply to every future request or just the ____ one.
What is the main practical benefit of a persistent instruction?
Free ai tools
Keep your streak going in Swipe AI
Swipe through every level of Prompt Engineering on your iPhone — with XP, stars, streaks, a personal advisor and the full track beyond these free levels.
Get the Swipe apps →