stonebox

Hiring hackathons for engineering teams

Find your next great engineer.

We run a hackathon built for your team. You hire the winner.

We build it around the real work of your open role. Candidates compete on their own schedule, with any tools including AI. Every submission is graded automatically on test data nobody sees in advance, so you get a ranked shortlist of who to hire without reviewing a single take-home.

Find your next great hire See how it works

Your first 5 candidates are free.

Stonebox · Live leaderboard

Senior ML engineer · hackathon

5 candidates, compared at the same minute

Hour

2:41

  1. 1Maya O.852
  2. 2Sofia R.740
  3. 3Jun P.694

An example hackathon leaderboard: five candidates for a senior ML engineer role, each shown at 2 hours 41 minutes into their own four hours. Maya leads with 852, then Sofia with 740 and Jun with 694.

Built for youFind people who can do your job
We build each hackathon around a problem like the ones your team solves, so the winner has already done the work.
AsyncNo scheduling, no drop-outs
Candidates start whenever suits them, so great people with day jobs take part. They still race on one leaderboard.
AutomaticZero take-homes to review
Every submission is graded on hidden test data. Your engineers only meet the finalists.
VerifiedAI welcome, no faking it
Candidates use any tools. A 5-minute voice interview about their own code shows who understands it.

The problem

A coding test can't tell you who can do the job.

  • They aced the coding test and couldn't ship.

    Puzzles measure puzzle practice, and AI now solves them in seconds. A pass tells you less every month.

  • Take-homes nobody has time to read.

    Your best engineers lose evenings to reviews, and every review is a judgment call you can't compare.

  • The best candidates drop out.

    Weeks of scheduled rounds lose the people with the most options, usually the ones you wanted.

The only reliable signal is watching someone do the real work. Stonebox lets you watch every candidate do it, without spending your team's time.

How it works

From open role to your next hire.

  1. Role
    Senior backend engineer
    Stack
    TypeScript, Postgres
    Problem
    Orders drop under load
    Scored on
    Orders, speed, accuracy

    1

    We build your hackathon

    Tell us about the role. We turn the kind of problem your team solves into a working codebase, hidden test data and a scoring rule. You approve it first.

  2. Time left

    1:19:07

    Practice score

    852

    Rank

    1/5

    2

    Candidates compete on their own time

    Each gets 4 hours, starting whenever they like, with any tools including AI. They submit as often as they want and watch where they stand on a live leaderboard.

  3. Final scoreUnderstanding
    1Maya Okafor85288
    2Sofia Reyes74074
    3Jun Park69431

    3

    You hire from a ranked shortlist

    Every submission is graded automatically on test data nobody saw in advance, and a 5-minute voice interview checks they understood their work. Spend your interviews on the top few.

The async hackathon

A real race, without the scheduling.

A hackathon is the closest thing to watching someone do the job: a real problem, a clock and a scoreboard. We made it asynchronous. Each candidate's clock starts when they press Start, and the leaderboard shows where everyone else was at that same minute. You don't lose great candidates to scheduling.

Scores at hour 2:41 of each person's 4 hours

  • Sofia

    Took it on Sunday, 2:10 PM

    740Finished at 812

  • Maya

    Taking it now, started 9:04 AM

    852

  • Jun

    Plans to start Thursday evening

    Not started yet

Maya is 2 hours 41 minutes in. Maya's leaderboard shows Sofia's score at 2 hours 41 minutes into Sofia's own run on Sunday, so it's a fair race even though they never overlapped. Jun will race them both on Thursday.

Why Stonebox

Not another coding test.

Coding tests check puzzles, take-homes eat your engineers' evenings, and AI interviewers judge how people talk. Here's how a hackathon built for your team compares.

  • Built around the real work of your role

    Yes

    Coding tests: no · Take-home projects: sometimes · AI interviewers: no

  • Candidates can use AI, like they will on the job

    Yes

    Coding tests: no · Take-home projects: sometimes · AI interviewers: no

  • Graded without your engineers' time

    Yes

    Coding tests: yes · Take-home projects: no · AI interviewers: yes

  • Every candidate measured the same way

    Yes

    Coding tests: yes · Take-home projects: no · AI interviewers: sometimes

  • Checks they understand the code they wrote

    Yes

    Coding tests: no · Take-home projects: sometimes · AI interviewers: no

  • A live competition candidates want to win

    Yes

    Coding tests: no · Take-home projects: no · AI interviewers: no

Built for your company

Every company gets its own hackathon.

Assessment tools hand you a library of generic tests. We build yours: the stack you use, a problem like the ones your team solves, and a score that rewards what you care about. The person who wins is the person who can do your job.

Example brief we'd build from

Role
Senior backend engineer
Stack
TypeScript, Node, Postgres
The problem
Checkout drops orders during flash sales
Scored on
Orders completed, p99 latency, no double charges
Time
4 hours

You get a working codebase with the problem built in, hidden test data, automatic scoring and interview questions tuned to the role.

Hackathons we've already built

Start from one of these, or we'll build one from scratch around your team's work.

See every example

What you get

Know who to hire before the first interview.

One line per candidate, ranked by how their work did on data they never saw, with how well they explained it.

Senior ML engineer · shortlist

5 candidates · fraud detection hackathon

Example
RankCandidateFinal scoreUnderstanding
1

Maya Okafor

Found the data bug in the first hour and explained every change clearly.

85288
2

Sofia Reyes

Careful, steady work. Strongest at making the service faster.

74074
3

Jun Park

Good practice score, but couldn't explain how their code worked.

Had trouble explaining their code		69431
4

Alex Novak

Fixed the crashes; ran out of time before improving accuracy.

54865
5

Dev Kim

Got a first version working but missed the main bug.

37158
Final score:
how their code did on test data they never saw.
Understanding:
how well they explained their own work in the voice interview, out of 100.

Maya is your hire.

Top final score, and Maya explained every change clearly. You've seen their real work before you've met them.

Jun would have passed your coding test.

A strong practice score, but Jun couldn't explain their own code in the voice interview. You find out before spending an hour in an interview, not after the hire.

Get started

Tell us who you're hiring.

We'll build a hackathon for the role and get back to you within a day to plan it. Then you send candidates one link each, and your shortlist builds itself.

Your first 5 candidates are free.

Questions

Will strong candidates spend 4 hours on this?

Yes, because it replaces the coding test and the take-home rather than adding to them, and it's a competition, not homework. They pick the time, use their own tools and AI, see a live leaderboard, and can finish early. Most engineers would rather show what they can build than solve puzzles on a call.

What is a hiring hackathon?

A timed competition built around a real engineering problem. Candidates get a realistic codebase, improve it, submit as often as they like, and are ranked by how well their work performs, so you see who can actually do the job before you spend any interview time. Ours are asynchronous: there's no event to attend, so each candidate takes part whenever suits them.

Do candidates need to be online at the same time?

No. Each candidate's 4 hours start when they press Start. The leaderboard compares everyone at the same point in their own 4 hours, so the race is fair whether they take part on Sunday morning or Thursday night.

What does "built for your company" mean?

We build the hackathon around your role: the language and tools your team uses, a problem like the ones you actually work on, and a scoring rule that rewards what matters to you, whether that's accuracy, speed, reliability or all three. You see it and approve it before any candidate does.

How is the work graded?

Automatically. Every submission runs against test data nobody sees in advance. While they work, candidates see a practice score; you rank them by a final score on separate test data, so tuning to the practice numbers doesn't help. The 5-minute voice interview at the end adds how well they understood what they built.

Can candidates use AI?

Yes, any tools they'd use at work, including Cursor, Claude and ChatGPT. That's how engineers work now. The voice interview asks about their own code, so you know who understands what they shipped.

How is this different from a take-home project?

With a take-home, your engineers read a pile of repos and judge them by feel. With Stonebox, every candidate works on the same problem, is scored on the same hidden test data, and lands on one ranked list. Nobody on your team has to review anything to get the shortlist.

How long does setup take?

We build your hackathon with you, usually within a few days. If one of our existing hackathons already fits the role, it can go out the same day.

What does it cost?

Your first 5 candidates are free. After that, we'll agree a simple price per candidate with you.