From notes to an argument
An essay is not a summary of what you read. This is the single most common misunderstanding in undergraduate writing, and it is why "you have summarised the literature but not argued anything" is the most common comment on returned essays.
An essay makes a claim and defends it. The literature is the evidence. Getting from twelve notes to a defensible claim is the hardest cognitive work in the assignment, which means — by lesson 1's rule — it is the part you must not delegate. It is also the part where AI is most useful, if you use it in one specific direction.
Find your claim in your reactions
Read the reaction lines of your notes end to end, ignoring everything else. Somewhere in there is a pattern: a thing you kept doubting, a disagreement you noticed between two camps, a gap nobody addressed, a claim you think is overstated.
Ananya's reaction lines kept circling one irritation: nearly every paper measured heat island intensity at night, because that is when the effect is largest and cleanest to measure, while the human cost of urban heat — outdoor workers, informal settlements, people without cooling — falls in the afternoon. The literature was optimised for measurability rather than for consequence.
That is an argument. It is arguable, it is supported by the pattern across her sources, and someone could disagree with it. It came from her own annoyance, recorded twelve times over three weeks, and no prompt would have produced it.
Use AI as the opposition, not the author
Once you have a claim, the most valuable thing available to you is a competent opponent.
"My essay argues that urban heat island research systematically underweights daytime effects because night-time measurement is cleaner, and that this skews policy. Give me the three strongest objections to this argument. Be specific and do not be encouraging."
The objections that came back were real: night-time effects have well-documented health consequences of their own, so "underweighted" needs to mean relative rather than absolute; daytime measurement has genuine confounds — direct solar loading is not the same phenomenon as the heat island — so the literature may be right to separate them; and policy in several cities does already target daytime exposure.
Every one of those had to be answered in the essay, and answering them is what made it good. The third one she had not known and had to go and check, which sent her to two more sources.
This is the correct shape of AI use for the hardest part of the work. You produce the claim; it attacks; you defend or revise. Nothing has been written for you and your argument is twice as strong.
Structure is a task, mostly
Once the claim and the objections exist, arranging them is much closer to a task than a skill, and it is fine to get help.
"Here is my claim, my three main pieces of evidence, and the three objections I must answer. Suggest two different structures for a 3,000-word essay, and tell me what each one emphasises and what it risks burying."
Two options with tradeoffs is a better request than one recommendation, because choosing between them is a real editorial decision that you should make and can only make if you see the alternatives.
Ananya's chosen structure put the objection about measurement confounds early rather than in a block at the end, because addressing it early made the rest of the argument possible rather than defensive. That was her call.
Write the first draft yourself
Badly, quickly, without stopping to fix anything. This is not a moral position; it is that the first draft is where you find out what you actually think, and a draft written for you skips the finding-out and leaves you defending a position you have not fully occupied — which becomes very obvious under questioning.
The practical version: give yourself two hours, work from your outline, leave gaps marked [check] rather than stopping to check, and do not read back what you have written until the end. What comes out will be poor. That is the correct result. Lesson 8 is about what you may legitimately do to it next.
Do this today: read your reaction lines end to end and write one sentence beginning "This essay argues that…". If you cannot yet, you need to read two more things — and knowing that is worth the ten minutes.