Reading as the reader

This is not a rules check. A rules check tells you the answer is 12 words over and contains a banned phrase. This tells you what a person will notice, feel, and decide.

Run it on the whole packet at once, not one answer at a time. Most of what this catches only exists across answers.

The three people

Read the packet three times, once as each. Who these three are depends on the type, so load application-types and read the matching reference/ file first. The roles below hold across types even though the people change.

The filter. Whoever decides you are worth more than a minute. A sophomore officer at 11pm with 200 left. A recruiter scanning against a requisition. A program officer checking eligibility before anything else.

Report: which answer would make them stop, where their attention drops, whether the first sentence of each answer earned the second, and whether anything made them uneasy in a way they could not articulate.

The specialist doing diligence. The person who actually knows the domain and will interview you if you pass. The officer who ran the project you mentioned. The hiring manager. The faculty member whose lab you named. They read for whether the claims hold.

Report: every number they would ask "how did you get that," every claim that reads bigger than it probably is, every place where the same fact appears differently in two answers or between an answer and the resume.

The person who has to live with the decision. Deciding whether they want to be assigned to a project with you for fifteen weeks, manage you for two years, advise you for five, or defend the grant when it is audited.

Report: do you sound like a person or a candidate, is there one moment of actual warmth or humor, does anything read as someone who cannot be told they are wrong, and would they be glad or tired to see your name.

That third reader is weighted differently by type. For student-org and grad-school it is close to decisive, because the reader is choosing a person they will actually work beside. For grant it is the weakest of the three, because they are funding a plan.

What to look for that a rules check misses

The same shape four times. Read every "why us" answer in a row. If they all run "here is what I built, here is what I lack, teach me," the reader notices the pattern even if each one is good. Vary at least one.

Weight with nowhere to land. A hard subject raised inside a word limit too small to resolve it. The reader gets the weight and none of the resolution, and closes the tab uneasy. This is a container problem, not a subject problem: same story, bigger question, works fine.

The unexplained hook. "After the year I just had." Points at something the reader does not know, and a stranger reads it as a warning rather than as depth. Say the thing or drop the reference.

Tonal whiplash. A joke immediately after something heavy. The fun questions come after the serious ones on most forms, and the reader's shoulders have just come down. Notice where the room changes.

Density of one theme. If one subject appears in five of ten answers, that is a texture observation, not automatically a problem. Reusing your strongest material across different orgs is what every applicant does and what orgs expect. Say what you noticed, give a read, and let the applicant decide. Do not argue that a reader going through 200 applications is cross-referencing one person, because they are not.

Answers that are too short. Anything under 60 percent of the limit next to answers at 95 percent reads as declining interest by the end of the form.

How to write the report

Numbered findings, most consequential first. For each:

1. What the reader notices, in their words.
2. Which answer, quoted.
3. Whether you would change it, and your actual recommendation. Not "consider whether."

End with the one change that would most improve the packet, and be specific about which answer and what to do.

When they push back

They know their life. If they say a finding is wrong, they are usually right, and the right response is to withdraw the argument in writing rather than restate it softer. If a piece of the finding survives for a different reason, keep that piece and say plainly which part you are withdrawing and why.