Have the AI unpack a clause I cannot read

You bought a critical-illness policy. Once a year. The first two years, on time.

Year three is tight. A month late feels fine — then three months vanish. You remember, call, and support says you can catch up. Pay, and the contract continues. Money is in. The confirm text arrives. You file it as closed.

A year-plus later, the company checkup finds something. Surgery. You pull the policy and count: bought in year one, now year four, waiting period 90 days — over three years gone. When you file, you feel solid.

The reply is a denial. Reason: after reinstatement, the 90-day wait restarts from that day. The claim day is day 47 after reinstatement.

You open the contract and find that passage. It was always there. You even glanced at it before you signed.

The waiting period of this contract is 90 days from the effective date or the most recent reinstatement date, whichever is later.

The gap between three-plus years and forty-seven days lives in this one sentence.

You know every character. The snag is reinstatement date. It has an industry definition, but it looks like ordinary words and never raises a hand to say it is jargon. Your brain files it as “the day it went live again, roughly,” and slides on. Whichever is later next to it does the same thing. It quietly picks the later of two start dates.

So after this sentence, what you keep is “waiting period, 90 days.” What it actually means is “waiting period, 90 days — and if you lapsed and paid back, those 90 days restart from the catch-up day.” The missing layer hides in the two words you thought you already understood.

This is the hardest kind of missing prerequisite: you do not know you are missing it. An obvious new word you look up. A definition dressed as everyday language never even makes you want to look.

Paste this at the AI and ask “what does this mean,” and you get a smooth explanation. You nod. You still cannot say why day 47 does not count. It explained the whole passage. You were missing one word. Explaining the whole thing is exactly how that gap gets covered over.

Ask another way, and it can do four things you cannot do for yourself.

1. Say how many things this passage covers, so you know which block to watch.
2. Unpack the sentence you are stuck on, list the concepts it assumes, and “reinstatement date” shows up on that list.
3. Fill one word at a time, analogize with something you already know, then quiz you.
4. Make you close the chat and go read the original sentence yourself.

The source is on the left. On the right: what you type at each step, and what it sends back.

The four steps differ only in how you ask. Swap the material below for whatever is in your hand, keep the order, and you have the flow you just watched.

Sentences one and two split the work: the map tells you which block to watch, the list tells you which layer you lack. Skip the first and jump to the second, and you can still walk. The cost is you do not know where this sentence sits in the passage, so you cannot plug it back in.

The half-sentence in number three — “do not go to the next word until I get it right” — is the key. Drop it, and the AI dumps all five words at once. You feel like you understand. Close the chat and only the last one is left.

Whether a word is filled depends on whether you can answer its quiz. Why those 47 days do not count — if you can say it, it is filled. If you cannot, go fill it again.

It gives you a map. The destination is the clause itself. Close the chat and reread the sentence that snagged you. That is the reading. This is a working habit. There is no experiment paper behind it.

What to ask before you start a forty-page thing: Have the AI tell me the five concepts I must know first. How to confirm you can say it after you fill: Let the AI play the outsider.