Why your job-search AI should tell you no
Every AI writing tool I tried during my own job search had the same failure mode: it was on my side in the least useful way possible. Ask it whether you're a fit for a role and it finds a way to say yes. Ask it to write the résumé and it rounds “I helped with” up to “I led.” It optimises for how the application feels when you send it, not for what happens after.
What happens after is the part that costs you. A flattering application gets you into interviews you can't survive, negotiating from claims you can't defend, in front of the small number of people you most needed to impress. The expensive resource in a job search isn't applications — it's credible shots at roles you can actually win. Spending them on inflated ones is worse than spending nothing.
So jobhunt runs under one rule, enforced everywhere a sentence gets generated: don't claim what isn't true. Not as a disclaimer. As architecture. Here's what that cashes out to in practice.
The fit analysis leads with your real case, then names what's thin
When you paste a job posting, Coach scores you against the actual posting — stored verbatim, not a paraphrase — and opens with your genuine, evidenced strengths for the role. Then it names the gaps plainly, so you know where to focus. If a true hard blocker exists — a licence you don't hold, a clearance you don't have — it's named honestly, last, as something for you to weigh rather than a door slammed shut. What it will never do is paper over a gap with positive framing, because you're the one who has to sit in the interview where that gap comes up.
Your red lines are hard constraints, not suggestions
During the profile interview you can set things you've explicitly said never to claim — a title you didn't hold, a technology you only brushed against. Those become hard constraints on every document the system produces. They're stored with your profile, injected into every generation, and they don't erode over time or under pressure from a tempting job description.
Nothing “shipped” without evidence
The most common inflation in technical résumés is the quiet promotion of work-in-progress to work-completed. Coach won't write “shipped,” “deployed,” or “led” unless your own account of your work supports it. If you built a prototype that was deprioritised before launch, that's what the record says — and honestly, “built a working prototype; the project was deprioritised” is a more interesting interview conversation than a vague “delivered” that collapses under one follow-up question.
A second set of eyes, adversarially
After a résumé or cover letter is generated, a separate fact-checking pass audits every claim in the document against your profile, your original résumé, and the job description. It flags anything it can't support — it doesn't silently rewrite your documents, because silent edits are their own kind of dishonesty. You see the flags; you decide.
Doesn't everyone else inflate?
Maybe. But the filters you're worried about — the tracking systems, the keyword screens — are looking for relevance, not adjectives. A résumé built around the actual keywords of the posting, drawn from things you actually did, clears the same filters without planting landmines in your interviews. Honesty and effectiveness aren't in tension here. The inflated résumé is the one that fails late, expensively, in person.
I built this because it's the tool I wanted when I was searching: something that would find my real shot at a role and make that case as well as it could be made — and that would tell me no when no was the true answer, while I could still do something about it.
An honest picture — strengths first, gaps and any real blocker named — beats a flattering one. You decide where to spend your effort. That's the product.