Tech Screener
A first-round test you can read.
Turn a job description into a timed technical test, send each candidate a single-use link, and read the result question by question: marks for multiple choice, written AI feedback on every answer, and a summary of fit against the role.
Made for Hiring managers, Engineering leads, Recruiters running technical roles
- Clean up an effect on unmount+2Multiple choice · medium
- Event delegation on a long list+3Multiple choice · hard
- Selector specificity−3Multiple choice · easy
- Debounce a search boxreviewCode · medium
Correct debounce, and the delay is configurable. The pending timer is never cleared on unmount, so a late call can update a component that has gone.
- Marks
- 2 / 6
- AI score
- 72
- Fit
- Suitable
Why it exists
A first technical round usually costs an engineer an hour per candidate, and most of that hour confirms what a short written test could have shown. So teams reach for a test, and run into the next problem: either it is a generic quiz that says little about the role, or it is a black box that hands back a single score with no way to see how it got there.
Tech Screener builds the test from the job description you are actually hiring for, from a bank of questions you can read and edit. Candidates sit it on a single-use link with a clock they cannot reset. When it comes back, every mark points to an answer, every piece of AI feedback sits beside the answer it is about, and the summary of fit is advice for the person who decides, never the decision.
From a job description to a shortlist.
Build the test
Paste the job description and the skills are picked out of it, matching questions are drawn from the bank, and AI writes new ones where the bank runs short. Or hand-pick questions from the bank yourself and set the duration.
Send a link
Each link is for one candidate and one attempt. The candidate registers with a name and email when they open it, and nobody else can use it after that.
Candidate sits it
One question per page with a navigator, answers saved as they type, and a clock set by the server. Reopening the tab does not reset it, and the test submits itself when time runs out.
Read the result
Every answer sits beside its correct option or rubric and the AI's feedback on it. A summary rates the candidate against the job description and suggests what to probe in the next round.
What it does
- Tests from a job description
- Paste the JD, choose how many questions, which types and which difficulties. The test is assembled from the bank first, so questions you trust get reused, and topped up with new ones when needed.
- A question bank that grows
- Questions are tagged by stack, difficulty and topic, and can be searched, previewed and edited. AI can draft a batch for any stack you name, and anything it writes for a test stays in the bank.
- Marks weighted by difficulty
- Multiple-choice answers are marked as they are saved. A hard question is worth more than an easy one, and a wrong answer on an easy question costs more than one on a hard question, so guessing does not pay.
- Feedback you can check
- Written and code answers get two to four sentences of AI feedback against your rubric. The feedback is advice for the reviewer and shows next to the answer it is about, so you can disagree with it.
- A profile across tests
- When someone has sat more than one test, their results are combined into one profile: strongest skills, gaps, roles they suit, and what to develop. Reports export to PDF for the hiring panel.
- Fewer shortcuts
- Copy, paste and right-click are disabled during the test, and leaving the tab brings up a warning. Admin actions such as issuing links and changing roles are kept in an audit log.
Details
- Question types
- Multiple choice · written · code
- Code answers
- JavaScript editor; reviewed, not executed
- Difficulty
- Easy · medium · hard
- Marking
- Easy +1/−3 · medium +2/−2 · hard +3/−1
- Duration
- 1 to 300 minutes, 60 by default
- Invites
- Single-use link per candidate, shared by you
- Access
- Admin accounts by arrangement
Questions
Does the AI decide who gets hired?
No. Multiple-choice marks are arithmetic. The AI writes feedback on written and code answers and a summary of fit, and both are shown as advice next to the answers they are based on. A person reads the answers and makes the call.
Is the candidate's code run against tests?
Not yet. Code answers are written in a JavaScript editor and reviewed against your rubric by the AI and by you. They are not compiled or executed, and they do not get a numeric mark.
How do candidates get the test?
You create a link for each candidate and send it however you already contact them. There are no automated emails. The link works once, for the first person who registers on it.
Can a candidate cheat?
It is harder. Copy and paste are blocked, the clock is held on the server, and answers after the deadline are rejected. There is no webcam proctoring, so treat it as a first round and follow up in conversation.
How do we get access?
Get in touch. Admin accounts are set up for each hiring team, along with a starting question bank for the stacks you hire for.