One scheduled run, or one press of a button. Postings arrive from the feeds, duplicates are dropped, mechanical gates remove what is out of scope on its face, a deterministic screen sets aside the rest, and only what survives is read closely and ranked. The whole shortlist arrives at once, before the working day starts.
A rejection made on a stated requirement quotes the line that caused it. A requirement the posting never stated is recorded as not checked, never as a failure. Disagreeing with a verdict is therefore possible, because the verdict is legible.
Every posting the screen marks as likely is read. Ranking orders the work; it never authorises leaving the tail unread. A run that could not finish reports itself incomplete, with the number left unread stated.
A feed that errors or returns nothing is reported as an error. It is never presented as a quiet morning with no new jobs — the failure mode that lets a broken search look like a slow market.
Every posting shows which of its stated requirements your own record does not support, with preferences kept separate from requirements. It quotes the line rather than giving a percentage, and it calls no model, so it costs nothing.
Pursue, reject or defer. Each correction feeds back into the brief, so a verdict overturned once tends not to need overturning again.
Judgement has to come from somewhere. Dawnlist reads existing curricula vitae, then conducts an interview: what the evidence actually supports, what is ambition rather than record, which roles disqualify themselves outright, and which are worth a look in spite of the title. The result is a fit brief and a background factsheet, both open to editing at any point afterwards.
Before the first real morning, a sample of live postings is fetched and the verdicts on them are shown for correction. Each correction rewrites the brief. No daily run starts until that calibration is passed, because the transfer of judgement is the product rather than a setup step.
Licensed job-data feeds, reached through a service of this publisher’s own rather than by each reader signing up to anything. Providers can therefore be changed or added without waiting for an application update, and no account with a job-data provider has to be registered or paid for separately. The feed is part of the subscription.
Alongside that, the job-alert emails already arriving in an inbox — from any sender at all — can be dragged onto the application, which reads the postings out of them. No account, no password and no connection to any mail provider is involved: those are files, handled by the person they belong to.
A draft states only what the factsheet supports. A gap renders as a visible placeholder rather than a plausible sentence, and a draft still carrying one cannot be exported as finished. Figures keep the qualifiers they came with or are dropped altogether; nothing is quietly compressed into a stronger claim than the record will bear.
Three documents are written this way. A covering letter for a posting worth pursuing. A follow-up where a reply has not come. And a brief before an interview, which is the one document allowed to be discouraging — where the record is thin against what a posting asks for, the brief says so beforehand rather than afterwards. All three arrive as drafts, and nothing is sent.
Drafts are written in the sender’s own habitual register, measured from messages that person already wrote and dragged onto the application in the same way as the alert emails. Style only — greeting and sign-off habits, sentence length, how often a hedge appears. No employer, no figure and no claim is ever taken out of a sample, and below a handful of samples the measurement reports itself unusable and a neutral professional register is used instead.
Dawnlist uses generative artificial intelligence to judge postings and to draft your messages. If it produces something inappropriate, offensive or plainly wrong, report it and it will be read. The same route is in the application, under Settings.
A report is written by a person, not sent by the application. Dawnlist transmits no document at any point — not a draft, not an assessment, not a curriculum vitae — and that does not change here. Describing what happened is enough. Anything quoted in a report is in it because the person writing it chose to put it there.
Every company pursued is tracked inside the application: the stage it has reached, an evidence log of what was actually sent and received, and the date of the next step. Follow-up timing is computed from evidenced contact rather than from a remembered date, never falls on a Monday or a Friday, and an out-of-office reply carrying a return date moves the next approach to a week after that return. One live opportunity per employer, so a second posting at a company already in play is adjudicated rather than duplicated.
There is no integration to configure and no external board to keep in step. Everything held can be exported to a spreadsheet or a structured file at any time.