There is a specific kind of tiredness that comes from rewriting your own life, over and over, for someone who may never read it.
It isn’t the rejections. Rejections are fine — you get used to them, and most arrive before anyone has read a word. It’s the pre-rejection labour: the quiet hour before you even hit Submit, spent reshaping the same history so that this particular vacancy thinks it’s found its perfect candidate.
That is the point it tipped for me. Not into anger. Into repetition.
What The Market Actually Asks For 🇦🇺
Nothing about my situation is unusual, which is the problem.
I’m in regional Victoria. One regional rail journey plus a tram is a workable commute; anything that needs rail, suburban rail, and then a bus or a taxi is materially worse. I don’t have a full driver’s licence. Remote is ideal, hybrid in the CBD is very good, and a five-day city office can be considered for the right role — but “outer suburban, on-site, five days” is not a realistic ask.
I want permanent full-time, not fixed-term. Contracts and probation periods are risk flags I track, not a price of entry. Salary floor around AUD 60–65k, better at 65–75k.
And there’s the degree. I don’t have a university degree. After repeated rejections from one particular institution, I wrote myself a rule so I’d stop burning applications on it: do not apply there again unless the role genuinely doesn’t require one. I keep that rule because I kept ignoring it.
So the search is narrow before it starts. And into that narrow gap goes a stream of vacancies, each of which wants to be told — in its own dialect — that I’m exactly the person it described.
The Treadmill 🏃
Here’s the part nobody warns you about.
Applying for a job is not one task. It’s a fork: every vacancy splits into its own version of you. The skills section reorders. The summary at the top gets rewritten so it uses their words — customer success here, client operations there, onboarding when it’s a SaaS shop. The bullet point about the enterprise account moves up when the job is enterprise, and back down when it isn’t.
Same facts. Different shape. Every single time.
Do that often enough and you stop seeing it as writing. It becomes a mechanical pass — find their terminology, mirror it back, tighten the whitespace, check it still reads like a human wrote it, export, send.
The Windows instinct, if you’ve ever job-hunted there, is Save As: CV.docx, CV_v2.docx, CV_v2_FINAL.docx, CV_v2_FINAL_new.docx. Four files, one of which is the real one, none of which remember why the changes were made. It’s the version-control problem people solve with filenames until they don’t.
I wasn’t there yet. I had one document and a lot of re-editing. That was the system: me, doing the same mechanical pass again, and being a different candidate each time.
Two things bothered me, and neither was the rejection rate:
- The labour is entirely front-loaded. You spend the hour before applying; everything after that is out of your hands.
- I was doing the least interesting part myself. Rearranging facts to fit a job description is exactly the kind of work I’d hand to a machine if it showed up in any other context.
So the question stopped being “how do I write a better resume” and became “why am I the one doing this at all?”
The Handover 🤝
The answer was: I shouldn’t be.
Not because the writing is beneath me — because I’m the wrong person to evaluate my own fit. When you’ve been searching a while, you start arguing your case. Every vacancy becomes something you can make work. That’s not analysis, that’s optimism with a deadline.
So I gave the job to ChatGPT, and — this is the part that actually mattered — I gave it rules before I gave it a single vacancy.
I didn’t want an assistant. Assistants agree with you. I wanted something that would look at a listing, look at me, and say no, and here’s why.
The first line of my instructions:
Act as an analyst, not a yes-man.
Then the terms:
When I give you a vacancy, I want a critical assessment of:
- exact role fit;
- realistic chance of landing an interview/job;
- salary;
- employment type and permanence;
- probation risk;
- company stability/redundancy/restructuring risk;
- workload and burnout risk;
- hidden sales/retention/upsell/quota expectations;
- commute/hybrid requirements;
- skills/experience gaps;
- degree/licensing requirements;
- long-term career value;
- whether taking the role is actually financially rational for me.
Challenge my assumptions where appropriate.
Do NOT become overly enthusiastic just because a vacancy superficially matches my CV.
That last sentence is the whole design. Everything else is a checklist; that line is the character.
For each vacancy it gives me what the company actually does, what the role actually involves, a match rating, the gaps, the red flags, the likely salary if they didn’t state one, the commute reality, and a blunt recommendation: APPLY / CONSIDER / LOW PRIORITY / SKIP.
No essay. A table.
Why “Not A Yes-Man” Took Me So Long ⏳
Because the yes-man version is comfortable.
There’s a real temptation in using AI for a job hunt to ask it things like “tailor this resume for the role” and receive back something glowing — a cover letter that says I’m passionate about their mission, a summary that reframes retail experience as strategic client engagement. It reads beautifully. It is also, almost always, flattery with a word count.
The failure mode isn’t that the AI lies. It’s that it’s agreeable — and agreeable is exactly what you want when the silence is starting to get personal.
So I front-loaded the discomfort. The instruction to challenge me comes before any vacancy does, so it’s a standing obligation rather than something I have to ask for when I’m feeling brave. It’s much easier to tell a machine to be sceptical of you on a Monday than to ask it to stop flattering you on a Friday.
I also gave it the boundaries it would otherwise have to guess at: where I actually am, what I actually have, what I will and won’t take, and — the one that stings — which doors are closed for now. A system that knows your constraints can rank. A system you brief fresh every time will happily rank everything as worth a shot.
What Changed 🪟
The Windows habit died with the re-editing, and I didn’t notice straight away.
On Windows, the file is the system. You keep the document open, you tweak it, you save a new name, and your history lives in a folder of near-duplicates that you scroll through hoping one of them is the good one. The tool never asks what you’re optimising for, because it doesn’t know — it just holds the bytes you last gave it.
What I built instead doesn’t hold a document. It holds rules about a person, and the document becomes an output.
That’s the actual shift, and it’s the reason this is a series rather than a tip: the resume was never the hard part. The hard part was admitting that my judgement about my own fit had quietly become the least reliable input in the process — and handing the judgement, not just the typing, to something that would argue with me.
Have you ever caught yourself editing the same document into five different personalities in one evening? Tell me I’m not the only one. 👇