There is a number on a screen: fifteen thousand dollars. Two terms left on a bachelor’s degree, maybe three. I have stood in a room shaped exactly like this one before, and the last time, I did not finish. This time I am not writing an appeal letter. I am building a machine.
Not one machine. Two.
The first one goes looking. It’s built to search every week for scholarships that might apply to someone like me — a non-traditional student, first-generation, carrying a GPA that would disqualify me from more than half of what it finds. It hasn’t run a single week yet; I’m still building it. But it does not decide in advance that I am not worth searching for. It’s built to search anyway, and then tell the truth about what it finds.
That sounds small until you watch what “telling the truth” actually requires. When it checks a scholarship’s application link and the link comes back dead, the easy move is to drop the row and move on — one less broken thing to look at. I told it not to do that. A dead link gets flagged, not deleted. A listing whose deadline has passed could just as easily be deleted to tidy the list; instead it gets archived, kept, marked as expired rather than erased. And when a scholarship states a minimum GPA I can’t actually verify against what’s posted, excluding the row outright would be the cleanest move. I built a branch for that instead — not a silent no, a flagged maybe, routed to a queue where I have to look at it myself.
None of this deletes anything. Every flag becomes a row I have to look at myself, on a screen, some week — a queue instead of a clean list. I built a system that refuses the one shortcut every other system defaults to: quietly making the inconvenient thing disappear.
I built it that way on purpose, because I know exactly what disappears quietly when a Black, first-generation, nontraditional applicant sits down to write a scholarship essay. Not the facts. The shape of the story.
I have read a decade of prompts that ask for a hardship, and I have watched what they reward. Either you are the redemption arc — I worked three jobs, I nearly gave up, and now look at me — or you are the passive victim, acted upon by circumstance until a scholarship committee’s generosity rescues you. Both shapes are lies by omission. Both turn a structural failure — a balance a financial aid office treats as a rounding error and a student treats as the difference between staying housed and not — into a morality play about individual grit or individual suffering. The story gets to be inspiring or it gets to be sad. It does not get to be true.
So the second machine — the one that will eventually draft essays instead of just finding them — was not allowed to exist until I had written down what it could never say. No redemption-arc framing. No passive-victim construction. No inspirational abstraction without a concrete, specific referent behind it. Nothing that could be lifted whole and reused by a different applicant with a different life, because if a sentence works for anyone, it is not actually about me.
Underneath that list sits the rule that matters most, and it runs in two passes — but only the frame is machine-made. The first pass does not draft. It assembles: the facts, the lever this scholarship’s own language points to, the ban list, the exact prompt, the exact limit — and then it stops, leaving the sentence itself blank, because nothing downstream is allowed to write that sentence for me. The second pass audits what I put in that blank myself: does every number and date I used trace back to a single canonical account of my actual life (not a flattering version, not a strategic one, the real one), does the length hold. Fail either check and the draft doesn’t get to leave the building. It goes to a queue. I read it. I decide.
That is the same rule as the dead link. Flag it, don’t drop it. Don’t let the inconvenient thing disappear — not the broken URL, and not the parts of my story too messy to fit the shape scholarship committees have been trained to reward.
Here is the part I want to be honest about, because the whole design is built on the belief that omission is its own kind of lie: the second machine does not run yet, and when it does, it will not run on its own. It has no schedule. It only starts when I start it, by hand, one scholarship at a time — the same refusal of autonomy that shows up everywhere else in this system, just moved up a layer: I don’t get to automate my way out of deciding what gets said about my life. Not one essay has been drafted. And here is the decision I actually ended up making, not the one I expected to have to make: no model reads my age, my GPA, my reasons for leaving the workforce. I built the brief and the fact-check in plain code — the kind that cannot train on anything, cannot forget anything, cannot fold my life into whatever it answers next, because it was never reading my life to begin with. It counts words. It traces numbers back to a single true account of me. That is all it is allowed to do. The sentence about my own hardship still has to come from me, typed by hand into a table, because I never built the pipe that would let it come from anywhere else. The rules came first. The writing was never going to belong to anyone but me.
I am telling you this while the balance is still fifteen thousand dollars and the terms are still two or three. Nothing is solved. I do not know yet if this works. What I know is that the last time a number like this stood between me and a degree, I dropped out, and the system absorbed that as evidence about me instead of evidence about itself. I am not interested in giving it that evidence again.
The machine that searches will not let a dead link vanish quietly. The machine that writes, when it finally runs, will not let my life vanish into the two shapes this genre has always been willing to pay for. Flagging the thing instead of dropping it is not a technical preference. It is the only posture I have left that isn’t lying.