The actual job hunt is mostly silence punctuated by replies. You send 30 applications, hear back from 4. Of those, 2 are auto-acks, 1 is a polite rejection, and 1 — if you're lucky — is a recruiter wanting to schedule. Miss that one for 48 hours and you're at the back of someone else's calendar.
What Aplyqk does in that silence
Every few minutes our reply-polling worker checks the Gmail thread of every application you've sent. New message? It pulls in the body and runs it through a classifier — one of seven labels: interview request, question, rejection, offer, general acknowledgment, autoresponder, or spam.
If it's actionable (interview, question, general, offer), our drafter writes a contextual reply in your voice. The drafter has hard rules baked in: never invent facts about you, never commit to specific dates or salaries, never use placeholders. The output is something you could send as-is — but you don't have to. Edit, send, or discard from your dashboard.
The interview path
When the classifier flags an interview request, a second agent runs to extract structured details: the recruiter's proposed time slots, the format (phone / video / in-person), the duration, the platform if mentioned. These show up on your Interviews page as a slot picker — pick one, hit Confirm, and the interview moves into your Upcoming tab.
Calendar integration is on the roadmap — for now, confirming locks the slot in our records but doesn't auto-create the calendar event. Most users tell us they prefer making that decision themselves anyway.
Why the human stays in the loop
Aplyqk does not auto-send replies. Even if the drafter's text is perfect, the user still clicks Send. This is by design. The downside of agentic systems isn't latency or cost — it's the moment one sends something embarrassing on your behalf. We'd rather lose 30 seconds of your time per reply than ship that risk.