The BTL Runtime Hackathon is decided. 42 projects. 72 hours. 40M tokens burned. We did not judge on vibes. Every repo was code-audited. Every live app was tested. Every runtime call was verified against the usage ledger. Not vibes. Receipts. Here are your winners. 1st place: DOCENT ($500) A six-engineer AI swarm that reads your repo, argues about what it finds, and remembers what changed the next time you ask. 900+ verified runtime calls across chat, embeddings, and streaming, all load-bearing. Code: github.com/IamHarrie-Labs… Try it: trydocent-production.up.railway.app Demo: youtu.be/bXbu2HVZuh8 2nd place: TRIBUNAL ($300) Six model roles (panelists, judge, claim extractor, verifier, skeptic) debate live through the Runtime, with failover when a route dies and every cost header on-screen. The highest verified score of the event. Code: github.com/knockknock1456… Try it: tribunal-z2jl.onrender.com Demo: youtu.be/-LYkG-QIzMc 3rd place: CRITCACHE ($200) Code review built on cache economics. Cold run vs warm run, with the receipts to prove the savings. The cleanest build we audited: every commit inside the window, zero red flags. Code: github.com/DiverseXL/crit… Site: critcache.vercel.app Demo: youtu.be/Pem0BKjkL4Y Best Use of Runtime: VERIFI ($100) An autonomous QA agent that browses your app and finds real bugs. Per-stage cost ledgers built from our headers, cache-aware retries, and a real browser vision loop. 6M+ tokens of verified workload. The Runtime was not decoration, it was the engine. Code: github.com/torpadeka/veri… Demo: youtu.be/wze8sQ3KGCw Honorable mentions, all code-audited, all real: readIQ, student RAG with vision OCR: github.com/muhaj-dev/Read… Vintage, verified-computation investing agent: github.com/AIINSTARAJ/Vin… Dispatch, cache headers as product data: github.com/Xconmax245/Dis… Crosscheck, extraction you can trust: github.com/PugarHuda/cros… flex, multi-model contract audit swarm: github.com/Techkeyy/flex HIVE, scoped memory with auditor veto: github.com/Kaypoops/hive Icaruz, cache-economics knowledge market: github.com/arko05roy/Icar… BtlRadar, cost-tiered security cascade: github.com/TheWeirdDee/bt… Helena Labs, incident copilot: github.com/Pavan-personal… How we judged: 30 points Runtime use, 25 usefulness, 20 execution, 15 creativity, 10 presentation. Then we verified everything. Cloned every repo. Probed every live app. Matched every claim against the ledger. Congratulations to every team that shipped. The next event is coming.
