Build Your Personal Assistant Operating System

2.2

The Correction Loop Turns Mistakes Into Reusable Rules

Every correction is a future rule waiting to be named

Your morning brief will get something wrong on the first run. It will flag a newsletter as urgent. It will miss a rescheduled meeting. It will summarize an email thread and lose the deadline buried in the third reply. These are expected failures, and each one carries a lesson the assistant can use permanently if you capture it in the right shape.

The correction loop is the mechanism that turns a one-time fix into a reusable rule. When you correct the assistant's output, you are doing two things: fixing today's result and teaching a pattern for every future run. The difference between these two outcomes depends on whether you save the correction as a named rule or let it disappear into a conversation you will never reopen.

A graphic teaching image showing a mistake becoming a correction, then a named rule, then a review checkpoint — A correction becomes durable only when it is saved as a named rule and reviewed later.

Three levels separate a quick fix from a permanent improvement

A one-off correction fixes today. You tell the assistant: that email from the newsletter service is not urgent, skip it. The assistant adjusts this run's output. Tomorrow, the same newsletter arrives and the assistant makes the same mistake, because the correction was not saved.

A reusable rule fixes the category. You tell the assistant: emails from newsletter services are never urgent; classify them as informational. The assistant saves this as a named rule with a date and a scope. Tomorrow, the newsletter is correctly classified without your intervention.

A rule review fixes the system. Once a month, you review all saved rules, retire the ones that no longer apply, resolve conflicts between rules that contradict each other, and confirm the ones that are still working. The rule set stays clean and current.

Three correction levels

Level	What you do	How long it lasts	When to use it
One-off correction	Fix this specific output	This run only	Rare errors you do not expect to repeat
Reusable rule	Name the pattern and save it	Every future run until retired	Recurring patterns: sender classifications, priority thresholds, format preferences
Rule review	Audit all saved rules	Refreshes the full rule set	Monthly, or when conflicting rules produce inconsistent output

A named rule has a date, a scope, and an owner

Every reusable rule should include four pieces of information: a plain-language name (what the rule does), the date it was created, the scope (which modules it applies to), and the source (which correction triggered it). Without these, rules accumulate silently and you lose track of why they exist.

Approval boundaries control what the assistant can do without asking

Every has actions the assistant should perform automatically and actions that require your explicit approval before proceeding. The line between these two categories is the . Drawing it clearly prevents the assistant from taking actions you would never have authorized while still letting it handle routine work without interrupting you.

The is a spectrum. At one end, the assistant can only read data and produce reports. At the other end, it can take actions on your behalf: sending emails, scheduling meetings, archiving messages. Most modules start at the read-only end and graduate toward more autonomy as you build trust through successful runs.

The approval ladder

Level	What the assistant can do	What needs your review	Assistant behavior
Read-only	Read sources and produce summaries	Everything else	Morning brief: reads calendar and email, produces a summary, takes no action
Draft-only	Read sources and draft proposed changes	All proposed changes before they take effect	Email : drafts replies for your review, never sends
Suggest-change	Propose specific actions with one-click approval	Each proposed action individually	Task : suggests rescheduling a low-priority task, waits for your yes/no
Low-risk autonomous	Execute safe actions automatically	High-impact or irreversible actions	Email : archives newsletters automatically, holds client emails for review
Restricted	Never autonomous for this category	Every action, always	Financial data: never sends, shares, or modifies without explicit approval

A graphic teaching image showing approval levels rising from read-only to restricted with review at the top — Approval boundaries are not all-or-nothing; each action type can sit on a different rung.

A 's approval level can differ by action type. The email might be autonomous for archiving newsletters (low-risk, reversible) and restricted for sending replies (high-impact, irreversible). Drawing boundaries per action type gives you fine-grained control without forcing the module into a single trust level.

Monthly rule maintenance retires what no longer applies

Rules accumulate. After three months of active use, a might have fifteen rules. Some address problems you solved by changing your workflow. Some conflict with rules you added later. Some apply to tools you no longer use.

A monthly rule review takes ten minutes. You list all active rules, ask the assistant to flag any that have not triggered in the past month, and retire the ones that no longer match your current situation. You also check for conflicts: two rules that give opposite instructions for the same scenario.

A graphic teaching image showing Rule A and Rule B pulling a classifier toward inconsistent output until a merge checkpoint resolves them — Contradictory rules do not fail loudly; they create inconsistent output until you merge or retire one of them.

Your turn

Run a monthly rule review

You'll review all active rules, retire stale ones, resolve conflicts, and confirm the rules that are still producing good results.

Why this exercise matters

Without periodic review, rules accumulate until the assistant's behavior becomes unpredictable. Stale rules produce unnecessary work; conflicting rules produce inconsistent output. Ten minutes of monthly maintenance prevents both.

You’ll leave with

A clean list of active rules with confirmed scope and relevance.
Stale rules retired with a note about why.
Conflicting rules resolved into single clear instructions.
A count of rules added, retired, and modified this cycle.

Use the prompt in order

1
List all active rules across modules
Ask the assistant to list every saved rule with its name, date, scope, and the last time it triggered. Sort by so you can review each module's rules together.
2
Retire rules that have not triggered recently
If a rule has not triggered in the past month, check whether the situation it addresses still exists. If the underlying problem was solved by a workflow change, retire the rule. Mark the retirement date and reason.
3
Identify and resolve conflicting rules
Ask the assistant to flag any pairs of rules that could produce contradictory instructions for the same input. For each conflict, decide which rule takes priority or rewrite both as a single clearer rule.
4
Confirm the remaining rules
For each rule that survives the review, confirm it is still accurate and useful. Update the scope if your modules have changed since the rule was created.

Starter prompt text

Open the full text if you want to check what will be copied.

Show starter prompt

List all my active assistant rules across every module. For each rule, show: name, creation date, scope (which modules it applies to), and the last time it triggered.

Then walk me through a cleanup:
1. Flag rules that have not triggered in the past month.
2. Flag pairs of rules that could contradict each other.
3. For each flagged item, ask me whether to retire, merge, or keep the rule.

At the end, give me a summary: rules kept, rules retired, rules merged, and the total active count.

What the answer should give you

A table of 12 active rules across 4 modules. 3 rules flagged as untriggered (1 retired because the sender no longer exists, 2 kept as safety nets). 1 conflict found between email urgency rules (resolved by merging into a single priority rule). Final count: 10 active rules, 2 retired, 1 merged.