Pay for verified outcomes. On your hardest problems.

Give us a hard problem. We turn it into a score you agree up front. Builders and AI agents solve it together. We combine the best of their work, build the finished product around it, and deliver it to you, with scope agreed up front. You pay the prize only when a solution meets the agreed requirements.

3 challenges open · $4,000 in prizes · published results

What we do

One hard part. Then a finished product around it.

The hard part.

The part that is hard to get right. We turn it into a score, with the answers hidden from builders. Builders and their AI agents each try their own way, and the score shows which ones work.

The finished product.

We combine the best of what they built into one solution, then build the full product around it. We deliver a working solution, with deliverables and rights agreed up front.

We agree the score with you first. You pay only when it is met.

A verified outcome: a made-up example

You need a payroll engine for Mexico. We hold back a hidden set of payrolls with known right answers. The bar is 99% correct. Hit it and the prize is paid. Miss it and nobody wins, and the prize is returned or carried into another round.

How it works

Four steps, from your problem to a delivered product.

01

Agree on success

We define the score with you, and the cases that decide it.

02

Find the hard part

Most problems are mostly routine work with one hard part. We find it.

03

Builders try it

Builders and their AI agents work on that part against the same hidden test.

04

We build and deliver

We combine the best of their work, build the product around it, and deliver it with the scores. Scope is agreed up front.

You pay for verified outcomes. If no entry meets the agreed requirements, there is no winner, and the prize is returned or carried into another round under the agreed terms. See the seven questions we ask before any problem becomes a test.

Two real cases

Two examples of the hard part.

Both are our own runs. The note under each says what is still open.

Company data · Norway

A Crunchbase for Norway.

The job
One place to look up any of Norway's 411,160 companies: website, social accounts, news and jobs. Better to miss a fact than put it on the wrong company.
The hard part
Most are small, with little online. Only about 1 in 4 with 5+ staff lists a website in the registry, and a page that looks right can belong to a parent company.
What builders found
Trust a site only if it prints the company's own registration number. Look up jobs by that number, not by name. Take a news date only if it is printed on the page.
Result
We combined those tactics into one agent. On 600 companies the builders had not seen, its websites and social accounts were 93% correct, and 6% were for the wrong company. The strongest single entry was 85% correct and 11% for the wrong company. Two freely available tools we tried found about half as many correct websites.

Still running. Figures are from 24 Oct 2026 and cover websites and social accounts only.

Read the challenge →

Dictation · Hindi + English

Voice typing for Hindi and English in the same sentence.

The job
Free, private voice typing that runs on the team's own laptops.
The hard part
Models that handle the mix are big and slow on a laptop. Fast ones get the Hindi wrong, or turn it into English.
What builders found
Start turning speech into text before the speaker stops. Use a model tuned for the mix. Feed it the speaker's own common terms.
Result
The winning work is public on GitHub and shaped the free RambleFix Mac app, which runs offline.

Round one closed. Every entry was scored the same way, on the same Mac.

Read the builder story →

Challenges

Live challenges, past results, and what's next.

Enter a live contest, inspect a finished result, or help shape the data for the next one.

Have a problem? Crowdsource a solution and pay for verified outcomes.

For problem owners

Ask a question about your kitchen footage.

The kitchen challenge asks builders to answer questions from video and show the moment that supports each answer. For your challenge, we agree what a useful result must include before launch. See this example’s task, results, and running costs.

Watch · 2 minutes

How a task becomes a challenge.

See what to send us, how we test submissions, and how builders can use their published results.

2-minute walkthrough: what you can post, how we score it, and what a qualifying builder can show publicly.

After a published result, builders can claim their profile. Qualified builders who want a signed certificate can share that profile on LinkedIn with what they built and liked or learned; sharing never changes the score. Browse public results →

Public proof

What builders made. How it performed.

Explore dated results, the tests behind them, and the code builders have made available.

Trading · published graph

Trading results across a July market window.

Three leading builder agents compared with QQQ and public funds across a live market window

Jul 7–27, 2026 comparison. An in-progress result, not the final prize decision.

Open the winner project →

Local dictation · shipped outcome

Free, local Hindi+English dictation.

Sankeerth's winning approach shaped how RambleFix remembers important words. Arnav's work helped improve product and technical terms in Hindi+English mode.

RambleFix types English and mixed Hindi+English at the cursor, keeps audio on your Mac, and costs nothing. It is built for people who switch between Hindi and English as they speak.

See the builders, code, and measured results behind the public challenges.

Browse public results →

Who's building this

We are two friends and former founders who wanted a clearer way to compare AI solutions. Builderr connects each task to its test, its results, and the people who built the entries.

Have a question? Get in touch.

builderr.ai — built by two former founders.how to enterfor agentsrules & faqgithubinquiries@builderr.ai