7

Systems, not sessions

Module 3 · Structure. After this lesson you can: write a standing instructions document that holds quality consistently across sessions without requiring daily re-setup.

A common pattern among people who use AI tools regularly: every session begins with a reset. The same preferences re-stated, the same warnings repeated, the same context rebuilt from memory. On a focused morning, the re-setup is thorough and the work is excellent. On a distracted morning, parts get skipped, and quality drops in direct proportion to what was omitted.

The pattern exposes a structural weakness. Quality that depends on what someone remembers to type each morning varies with their mornings. Standards that live only in a person's memory are subject to everything that memory is subject to: fatigue, interruption, competing demands. The AI's context window is well-understood to have limits; the limits on human recall at session start are less often discussed, but equally real.

The fix follows directly from Lesson 5: facts in a poor location belong in a better one. The rules built through Lessons 3 through 6 all belong in a single plain document the AI reads at the start of every session: verify before confirming, check the source, maintain one authoritative location for facts, diagnose root causes before patching symptoms. Most AI tools provide a designated place for this kind of document. Different tools call it by different names: a system prompt, custom instructions, or a profile. The general term for this document is standing instructions: rules the AI reads at the beginning of every session, regardless of what the operator remembers that day.

Once standing instructions are in place, session-to-session variation largely disappears. A focused-morning session and a distracted-midnight session receive the same baseline quality, because neither depends on the operator's recall. When the system produces a new kind of error, the appropriate response is not just to correct it. Apply the Lesson 6 loop, identify the missing rule, and add one line to the document. The standing instructions grow at exactly the places where real use has scraped them.

Over time, a well-maintained standing instructions document shifts the dynamic measurably. The system requires less direct operator input each month, not because the AI has changed, but because every rule learned through experience is encoded in the document and runs automatically. Sessions evaporate when a conversation closes. Rules in a standing instructions document accumulate.

A practical test for any process: does it need less involvement from the person running it each month? If yes, it is functioning as a system. If it needs the same amount of involvement or more, it is a habit wearing the name of a system. The purpose of building standing instructions is to make that shift.

Stop here. Sit with this before continuing.

What do you re-explain every time, and why is it not written down?

Try this nowUnder 30 minutes
  1. Open the Shelf and copy the starter constitution. It is five rules, each one introduced in an earlier lesson.
  2. Rewrite each rule in your own words. Word-for-word copies do not stick. Rules phrased in your own language are more likely to be followed and extended.
  3. Add two personal rules. Think of the last two times an AI produced something frustrating. Each frustration is a rule not yet written.
  4. Put it where the machine sees it every time. Most AI tools have a place for standing instructions or a profile. If the tool in use does not, paste the document as the first message of every session until it does.
  5. Live with it for a week, then edit. The first version is a draft. The editing, rule by rule, scrape by scrape, is the actual practice.

Two questions before you go

Answer, then say whether you were sure or guessing. Being honest about which is the skill being trained.