TL;DR: “One more prompt” is one of the most expensive phrases in builder culture. This framework gives you clear, objective criteria for when to close the laptop — so you’re not making that decision inside the session, where bias and momentum always win.

The Short Version

You know the feeling. It’s late. The work is going reasonably well. You could stop — you’re not sure you should stop — and there’s one more thing you want to check. And then there’s the next thing. And an hour later you’re still at the laptop and the session quality has been declining for 45 minutes but you haven’t noticed because you’re inside it.

The problem isn’t discipline. The problem is that you’re trying to make the “when to stop” decision inside the session, where every incentive says to keep going. That’s not when the decision should be made.

Why “Stopping” is a Decision You Can’t Make Well Inside a Session

Decision fatigue research gives us a clear picture: the quality of decisions degrades over the course of a decision-making period. After a long session of choices and direction-setting, your capacity to make the next decision well is meaningfully lower than at the start.

“Should I keep going?” is a decision. And it’s being made exactly when your decision-making capacity is lowest — late in a long session, when you’re tired and invested and have momentum.

📊 Data Point: Research on judicial decision-making found that favorable parole decisions dropped from 65% to near 0% over the course of a judicial session, restoring to 65% after a break. The decision changed; the cases didn’t. Decision fatigue was the variable.

The same effect happens in your head when you’re trying to decide whether to close the laptop.

💡 Key Insight: Make the stopping decision before the session starts, not during it. Rules set in advance override the in-session bias toward continuation.

The Framework: Five Criteria for Closing

Use this framework before each session: pick the criteria that apply, and commit to stopping when the first one is met.

Criterion 1: The Timer

The simplest criterion. Set a time limit before you start. When the timer goes off, close. No extensions.

This is the most reliable criterion for most people most of the time. Its reliability comes from its absoluteness — you committed to it before the session, so the in-session brain doesn’t get to renegotiate.

Criterion 2: The Defined Output

Before you start, define the specific output you’re working toward. When that output exists — not when it’s perfect, when it exists — close.

This is the quality criterion for task-focused sessions. The risk is perfectionism: “it exists but it could be better.” The rule is: it exists and is good enough. The standard for “good enough” is defined before the session, not inside it.

Criterion 3: The Quality Signal

Your body knows things before your conscious mind admits them. These physical signals indicate declining cognitive quality:

  • Shoulders hunched, neck tight
  • Eyes dry, blinking less frequently
  • Prompt quality declining (you’re rambling, repeating yourself)
  • You’ve read the same AI response twice and can’t retain it

When two or more of these are present: close. Not because you’ve failed. Because continuing is producing output you’ll edit or discard tomorrow anyway.

Criterion 4: The Time of Day

A simple rule that doesn’t require any in-session judgment: no AI after [specific time]. Whatever time that is for your life — 9pm, 10pm, 11pm — it’s a rule, not a guideline.

This criterion works because it removes the “how am I feeling” variable entirely. The time is the time. The laptop closes when the time comes.

💡 Key Insight: Rules that don’t require judgment are more reliable than rules that do. “Close at 10pm” is more reliable than “close when I’m tired” because you don’t have to evaluate your own state at 10pm.

Criterion 5: The Relationship Check

If someone in your life — a partner, a friend, a family member — is waiting for you, present in the same space, and you’re still in an AI session: that’s your cue. Not in five minutes. Now.

This is the relationship protection criterion. It converts an abstract principle (“relationships matter”) into a concrete observable: a person, in your presence, whose time is more important than the next prompt.

Building the Pre-Session Ritual

The framework only works if you apply it before the session. The pre-session ritual:

  1. Set your stopping criterion. Pick one or more from the framework above. Write it down (physically) before you open any AI tool.

  2. Prepare the environment. If you’re using a timer criterion, set the timer before you open the tool. If you’re using a time-of-day criterion, set a phone alarm for that time.

  3. Define your output. Even for exploratory sessions: “I’ll explore this for 20 minutes and capture 3–5 insights.” That’s a defined output.

  4. Name the override exception. Is there any condition under which you’d extend the session? Name it in advance. “If I’m mid-sentence when the timer goes off, I’ll complete the thought and then close.” The exception is narrow and specific, not “if I feel like I’m in flow.”

The Post-Session Close

When the criterion is met: a specific physical action to close the session.

Close the laptop (not minimize, close). If you can’t close it: close every AI tab and the browser. Turn on a different room’s light. Make a brief note of where you are and what the next step is.

The physical action matters because it creates a sensory boundary. Your body knows you’re done. The ritual reinforces the decision the rule made.

What This Means For You

The “one more prompt” loop isn’t a discipline failure. It’s a decision architecture failure. You’re trying to make a biased decision with the wrong tools at the wrong time.

This framework moves the decision to before the session, where you can make it clearly, and then automates the execution with rules and timers that don’t require willpower.

Key Takeaways

  • The “when to stop” decision should be made before the session, not during it — in-session bias always favors continuation
  • Five criteria: timer, defined output, quality signals, time of day, relationship check
  • Rules that don’t require judgment (time-of-day) are more reliable than rules that do (how am I feeling)
  • A physical close ritual (close the laptop) reinforces the decision with sensory confirmation

Frequently Asked Questions

Q: What if genuinely important work comes up after my stopping criterion is met? A: Define “genuinely important” in advance. Customer emergency? Yes. Interesting idea you want to develop? No — that’s what tomorrow’s session is for. The pre-definition prevents the in-session rationalization that turns any interesting thing into “genuinely important.”

Q: Do I need to use all five criteria? A: No. Pick one or two that fit your situation. The timer alone, applied consistently, is sufficient for most people. Add others as you learn what your specific patterns are.

Q: What about weekend or personal AI use — do the same rules apply? A: The same principles apply, with different specific criteria. Personal AI use also benefits from pre-defined sessions and stopping criteria. Recreational exploration can have a longer timer (30–60 minutes) while still being bounded rather than open-ended.

Not medical advice. Community-driven initiative. Related: Time-Boxing AI Sessions | AI-Free Hours Protocol | Relationships vs AI Time