threshold

Fetches doctoral positions, scores them against your resume, and drafts what each one asks for.

Email Password

Display name optional

threshold

At most new positions

Fetch log

Order Tier Not yet applied Hide the ones on hold

Pick a project on the left.

A project is a resume plus the filters derived from it. Press Fetch positions and everything the site has published under those filters lands here, scored.

Applied on the source site On hold

Why it is on hold

Open the advert Application form

Score

What this application asks for

The post

Activity

Version

Download PDF Download text

No documents generated for this position yet.

The buttons above are the ones this advert actually asks for. Press one and the draft appears here.

Facts it needs from you

What changed in this version

Comments I could not apply

Conversation

Every turn below is about this document alone. The next revision is given all of it, along with the log of the runs that wrote these versions, so a short reply — "shorter than that", "no, keep the solver paragraph" — means what it would mean to a person.

Your comments on this version

Say what is wrong and where. "Paragraph three reads like a CV summary, cut it to one sentence about the solver" gets you further than "make it better". Anything you do not mention stays untouched.

New project

Upload your resume. It is read to work out which of the site's 7,600 categories your work falls into, and those become the project's search filters.

Choose a PDF, DOCX, TXT or MD file

Reading it…

Project name Track

What it found in your resume

Categories it will search

These are ORed on the site, so more categories means more results, never fewer. Untick anything irrelevant; the default selection is deliberately wide.

Add a category the reading missed

Settings

Access to the OpenAI API

Checked in layers, so a failure names the layer that broke rather than saying "it doesn't work".

Also try public DNS resolvers if the name does not resolve

Public resolvers that answered

API key

Stored in the database for this installation, verified before it is saved. Only its first six and last four characters are ever shown back to you.

Models

Fetching and scoring Drafting documents

Password

Backup, and moving between installations

This is for two things: keeping a copy of what you have collected, and not collecting it twice.

The source site allows a limited number of fetches a day and every score costs a model call, so the adverts in this database are expensive. A bundle is one file holding them: the adverts, their scores, what each one requires, and optionally the documents drafted against them. Download one as a backup, or to carry your search to another machine. Importing is a merge — nothing is deleted, and importing the same bundle twice changes nothing — so it is safe to do whenever you are unsure.

It is not the database backup. For that, run docker compose exec db pg_dump -U threshold threshold, which also captures your accounts and settings.

Export

Include generated documents Include resume text

Import a bundle

Choose a bundle file

Create projects that do not exist here Include documents

Or pull straight from the other installation

Needs SOP_SYNC_TOKEN to be the same on both sides. No file in between.