You upload a resume once. It is read into a list of separate facts — each job, each project, each skill — and you confirm them. From then on, every resume it writes for you is assembled out of facts you confirmed, tailored to one job description at a time.

It is built so that it cannot, and the check is in code rather than in the instructions given to the model. Every sentence it writes has to cite the confirmed facts it rests on; anything citing a fact you do not have is thrown away and rewritten. Numbers get a second check — any figure that does not appear in a fact you cited is rejected, so a percentage or a team size cannot be conjured up to fit a posting.It can still choose emphasis badly or word something in a way you would not. Read what it writes.

Every feature needs it, which is stricter than most products. Until the address is confirmed, nothing runs. If the message does not arrive, ask for another one on the confirmation page — that link is on the login screen too.

Credit values are configuration placeholders for the future billing model. The current build records AI token usage but does not decrement credits or block an action when a balance reaches zero. The provisional costs are:

• · Reading an uploaded resume — 3
• · Breaking a job description into requirements — 2
• · Generating a tailored draft — 8
• · Asking for a revision — 2
• · Translating a draft — 6

The planned free allowance is 20 credits, valid 3 days from the moment you confirm your email. These values may change before billing is enabled.

Yes, directly, and it saves as you type. You can also ask for changes in words — “make it fit one page”, “more backend emphasis” — and the reply comes back as a list of proposed edits with the reason for each. Nothing changes until you accept it.

Nothing is rewritten behind your back. Drafts you already generated keep saying what they said, and the editor tells you when one of them rests on a fact you have since changed or removed, so you can regenerate if you want to.

English is the supported product language today. The data model keeps interface, Profile-content and Resume-output locales separate so Simplified Chinese can be added without rewriting stored projects, but the complete Chinese experience is not released yet.