Fun stuff · entry 02

Hire or Nah

Jev is basically a model for making judgement calls. Give it a job description and a resume, and it picks one of five verdicts from Strong Hire to No Hire. It also scores the match on skills, seniority, domain fit and whether the person has actually shipped anything.

Input 1 of 1 · job description

Paste any role. Jev compares it with my resume and gives you its honest call, even when the answer is no.

0 / 12,000 characters

I do not save your text. It goes to TypeSafe's Jev API, so please do not paste anything sensitive.

About the model

So what is a System One model?

Most language models generate an answer one token at a time. Jev only makes decisions from options you define up front. For this page, those options are the five hiring verdicts. It returns a probability for each one together with the smaller checks below, so it cannot suddenly go off and write a motivational speech instead.

Jev is the first public model from TypeSafe AI. They report response times of 70 to 500 milliseconds for System One tasks and do not charge for output tokens. The result here takes a little longer because I liked the fake deliberation animation too much.

Then Laya showed up a few days later. It is an open-source alternative built from research its creator published in March 2025, and it can run on a free Colab GPU. Apparently the race to build the world's fastest boring decision-maker is now a thing.

Jev docs Laya on GitHub