Winning code and benchmarks
This is what people are building on builderr: public submissions scored against real market data, not toy demos. Arnav's Round 1 agent had more upside when the Nasdaq market (QQQ) bounced, then stayed positive when it rolled over and finished down. Local dictation is the same pattern outside markets: a separate public board, 10 completed entries, leading submissions ahead of the open-source baselines we tested, and RambleFix as the higher prize bar.
That is the signal for anyone browsing: anyone can be a builder. Smart builders and students can build serious agents when the problem is public, the rules are clear, and the leaderboard is scored in the open. Trading has a published winner; Local dictation shows another benchmark being worked on in public.
Finalized at the July 2, 2026 close. On Jun 15, the Nasdaq market tracker (QQQ) had recovered to +0.2%; Arnav was already +12.7%. By the close, it had rolled over to -3.9%; Arnav was still +5.9%. He finished first ahead of Aarya and Robert using a leader-momentum strategy with regime filters, stops, position caps, and a measured leveraged ETF sleeve.
Round 2 opened July 7, 2026 and runs through the September 4, 2026 close. The benchmark is Arnav's Round 1 agent, now published for learning. Entrants compete against that same benchmark on a fresh forward market window.
The local dictation round is not won yet. On the public sample board, the leading submissions beat the open-source speech-to-text baselines we tested on Hindi+English clips. The prize bar is higher: beat RambleFix on the hidden same-machine run.
Builder profiles
Each board entry has a public, claimable profile. It shows what they built, and the public result. Qualification comes from the published bar, not from sharing. A certificate is available only after qualification; sharing the profile is optional.
Your build and reviewed result in one link.
A clear challenge bar decides eligibility.
Downloadable after qualification; sharing stays optional.
Built a Hindi-English dictation engine that isolates on-device inference and keeps a warm fallback ready if it fails.
All eight clips returned a final and this run cleared the current benchmark; two mixed-language clips remain the gap.
Built one local speech path for both rolling drafts and finals so Hindi-English wording stays consistent.
The revised entry returned every final and is close to the benchmark, but three clips were capped.
Built a small-footprint local engine that routes mixed speech and uses a stronger final pass without sending audio online.
All clips returned a final; mixed Hindi-English accuracy and slower finals remain the gap.
Built a private, local speech-to-text entry for the Hindi-English dictation challenge.
The latest revision returned every final, but four clips were capped and it remains below the benchmark.
Built an offline speech engine for ordinary CPU hardware, with a stable rolling transcript while someone speaks.
Two English clips beat the benchmark; mixed Hindi-English finals still missed too much.
Built a lightweight live draft path and a stronger on-device final pass after the speaker stops.
Several finals were strong, but one clip failed and the slowest finals lost ground.
Built separate local paths for English and mixed Hindi-English speech, then routes each utterance to the better fit.
Every clip finished, but several finals changed key meaning and arrived too slowly.
Built a fast local draft with a stronger Hindi-English final and bounded fallbacks when a model runs long.
One clip returned a usable final; the other seven did not finish with usable text.
Not on a board yet? Choose a challenge and build an entry →
On Jun 15, the Nasdaq market tracker (QQQ) had recovered to +0.2%; Arnav was already +12.7%. By the July 2, 2026 close, it had rolled over to -3.9%; Arnav was still +5.9%. The winning code is public so builders can study what worked.
| # | Agent | Account | P&L |
|---|
Refreshed from the latest fetched market bars. Round 2 prize and builder points require beating Arnav. Each agent starts a $100,000 account at its first scored market session.