Open challenge · $2,500
Signalpost company research
Build an agent that finds company information online.
Read Signalpost scoring →Rules
One short rulebook. Exact rules on each challenge.
These rules apply across Builderr. Each challenge page is the source of truth for its task, deadline, prize, scoring, benchmark, budget, eligibility, revisions and submission format.
Platform-wide
Rules shared by every challenge
01
The challenge page controls
Read it before building. If a global summary conflicts with the challenge page, the challenge page wins.
02
Same method inside each challenge
Entries use the same published scoring method, execution rules and minimum requirements for that challenge.
03
Submit an exact version
Provide the repository, commit, artifact or endpoint requested. The evaluated version—not a later replacement—is the entry of record.
04
Stay inside the stated boundaries
Respect the data, network, runtime, cost, model, source and revision limits listed on the challenge page.
05
Show honest evidence
Do not fabricate outputs, sources, runs or claims. Missing evidence should remain missing rather than being guessed.
06
Results are provisional until verified
A leaderboard can update quickly. A prize, winner status or certificate requires the exact verification stated for that challenge.
Exact scoring
Rules for each challenge
2 challenges are listed as open, with $3,500 in prizes. Use the relevant page below for the complete, current rules.
Open challenge · $1,000
Trading agents
Build a trading agent that buys and sells US stocks with paper money.
Read Trading rules →Published deadline passed · $300
Kitchen CCTV monitor
Build an agent that watches kitchen video and answers questions about what happened.
Read Kitchen CCTV scoring →Past result · $500
Local dictation
Build a local dictation tool that turns Hindi+English speech into text. Round 1 is complete.
View Local dictation result →What can invalidate an entry
- Failing a mandatory requirement or breaking an explicit rule on the challenge page.
- Using forbidden information, evaluator internals or another entrant's private work.
- Fabricating evidence, results, identities or eligibility.
- Malicious code, credential access, evaluator tampering or deliberate service abuse.
A low score, simple approach, honest bug or failed experiment is not misconduct. Builderr states the issue and gives the builder a chance to respond before a final discretionary call.
Code, privacy and ownership
- Never commit API keys, passwords or private data.
- Builderr uses submitted access under the challenge’s published evaluation, verification and sharing terms.
- You keep ownership unless separate written licensing or transfer terms were agreed before entry.
- Public result pages distinguish measured facts from private or unavailable artifacts.
Before payment
Winner verification
For challenges that require an exact scored agent, a provisional winner must provide read-only access to that exact agent version. Builderr checks reproducibility, eligibility and material rule compliance before issuing a prize, winner status, or certificate. An entry that cannot be reproduced or audited is prize-ineligible, and the next eligible entry moves up—or the prize remains unawarded— under that challenge's published policy.
FAQ
Where are the exact rules for a challenge?+
On that challenge's page. It is the source of truth for the task, deadline, prize, scoring, benchmark, budget, eligibility, revision limit and submission format.
Does every entry get the same test?+
Every entry inside a challenge is evaluated under the same published scoring method and execution rules. The inputs may be hidden, live or randomly selected when the challenge page says so.
Do I have to publish my code?+
Not unless the challenge says so. The submission page states whether public repositories, private read-only access or hosted endpoints are accepted. Never commit secrets to a repository.
Who owns submitted work?+
You keep ownership unless separate written terms are agreed before entry. A prize, certificate or public result does not silently transfer ownership.
Can a result be disqualified?+
Yes, for a stated rule violation, fabricated evidence, evaluator tampering, malicious code, identity abuse or other clear bad faith. A low score, simple approach or honest failure is not a disqualification.
When is a winner final?+
Leaderboard positions are provisional until the exact submitted artifact passes the verification stated on the challenge page. If it cannot be reproduced or audited, the entry is prize-ineligible and the next eligible entry moves up—or the prize remains unawarded—under that challenge's policy.
What if nobody clears the bar?+
Builderr does not force a winner. If no entry clears the published qualification bar, the challenge page's rollover or return policy applies.