Chapter 6 · Never Solve the Same Problem Twice
Never solve the same problem from zero twice.
A result ends the task. Residue lowers the cost of the next one. Good work solves today's problem; great work makes tomorrow's version cheaper.
Every solved problem either disappears or leaves something behind.
You answer a customer question. Tomorrow another customer asks the same thing, and you answer again. Nothing accumulates. Or you answer it and save a strong response as a template, and the next answer starts closer to finished. You fix a bug, or you fix it and add a test that prevents it coming back.
Call what remains residue: a checklist, template, decision rule, reusable prompt, rubric, test, documented mistake, workflow or automation.
TASK → WORK → RESULT + RESIDUE
First time, second time, third time
A simple habit:
- First time: solve it.
- Second time: notice it.
- Third time: systemize it.
The numbers are not sacred. The awareness is. A repeated problem is trying to tell you something. If you solve it manually forever, you are choosing to rent your own memory.
Standards beat prompts
People save prompts because prompts are visible. The higher-leverage artifact is usually the standard behind the prompt. A prompt says Rewrite this to be concise and professional. A standard says:
For our executive updates: maximum 250 words; the first sentence states the decision required; every number includes a comparison; no adjectives that imply performance without a metric; end with owner and date.
The standard survives model changes. The prompt is just one interface to it. Capture standards first, then let tools translate them into whatever the current system needs.
The Systemization Ladder
Move through increasing levels of structure, one rung at a time:
EXAMPLE → TEMPLATE → CHECKLIST → DECISION RULE → WORKFLOW → AUTOMATION → AGENT
Start with an example that shows what good looks like. Turn it into a template by asking which parts repeat. Write a checklist of what must never be forgotten, then decision rules for what changes case by case. Build a workflow for the sequence. Only then automate the stable steps, the ones software can do without judgment. Use an agent only when the task genuinely requires choosing intermediate actions inside a defined goal and boundary.
Do not skip rungs because "agent" sounds impressive. Complexity has a cost. The simplest reliable system usually beats the most fashionable one.
Draw the escalation line
Draw an imaginary line through every automated system. Below the line, the machine acts. Above it, a human judges. Where should it sit? Ask four questions:
- Consequence: how costly is a wrong action?
- Reversibility: can it be undone easily?
- Uncertainty: how sure is the system about what is happening?
- Context: is important information likely to sit outside what the system can see?
Low-consequence, reversible, familiar actions can sit low. High-consequence, irreversible, uncertain decisions move up. Automation does not remove judgment. It moves judgment into where you draw this line.
Five parts of a reliable system
Whether it is a checklist or a sophisticated agent, define five things:
- Trigger: what starts the process?
- Context: what information does it need?
- Standard: what counts as good?
- Action: what may the system actually do?
- Escalation: when must it stop and ask a human?
If you cannot define these, you may be automating confusion.
Don't automate mistakes
The safe order is experience, extract, systemize, automate, evaluate, improve. Automating a process whose exceptions you do not understand just makes the mistakes faster. Run it by hand long enough to know where it breaks.
Systems should create silence
A good system does not just produce more. It removes noise: fewer repeated questions, fewer decisions made from scratch, fewer things that need you. That quiet is the compound interest of operational work.
Never solve the same problem from zero twice.