RuleBound: The Sealed Build Challenge

You will not see the exact problem until registration closes. Every registered participant receives the same brief, synthetic data and starter kit at 10:00 AM IST on 25 August 2026. Then the 48-hour clock begins. About the challenge: RuleBound tests the boundary between creative AI and deterministic software. Build a system that can generate useful structured output, enforce hard constraints, explain its decisions and produce repeatable results. The detailed problem remains sealed until Round 1 begins. Who should participate: Backend and systems engineers Builders familiar with rules engines, pricing engines, schedulers, constraint solvers or compilers AI engineers working with structured output, tool use or constrained generation Developers with CAD or spatial-format experience such as DXF, DWG or IFC Final-year students, recent graduates and professionals with substantial technical projects You do not need every listed skill. A working portfolio and the ability to defend your architecture matter most. Format and schedule (all times IST): Registration: 18–24 August 2026; closes 24 August at 11:59 PM Round 1 — Sealed Sprint: 25 August 10:00 AM to 27 August 10:00 AM; every valid registrant participates Round 2 — Live Technical Defence: 28 August; top 8–10 receive 25-minute screen-share slots from 2:00 PM Round 3 — The Final: 29 August 10:00 AM to 31 August 10:00 AM; four finalists Winner announcement: 1 September 2026 Participation: Online. Individual participation only. Teams and joint submissions are not permitted. Everyone may apply. Round 1 deliverables: Repository runnable with one documented command ARCHITECTURE.md, maximum two pages, including a section titled “Arbitration” Demo video, maximum five minutes, showing a deliberate constraint violation and its repair Committed output for every released specification Determinism statement with the exact repeat-run command Scoring: Arbitration — 35 points Price integrity and determinism — 25 points Constraint enforcement — 15 points Generation quality — 10 points Engineering quality — 10 points DXF/DWG ingest or export — up to 5 bonus points Azure deployment with Entra ID — up to 5 bonus points Important judging notes: UI polish earns zero points. A terminal and static plan view score the same as a polished interface. No language model may operate inside the deterministic pricing path. Prize and internship opportunity: Winner: ₹10,000 cash prize Up to two finalists may receive ₹50,000/month paid internship offers, subject to final evaluation, technical defence, demonstrated authorship, availability and role fit. Ranking alone does not guarantee an internship. Rules: All challenge work must begin after the Round 1 reveal. Submissions must be original or properly licensed. AI coding assistants are allowed, but participants must understand, defend and modify their submission live. The repository, documentation, video and links must remain accessible throughout judging. Attempts to obtain or share the sealed brief early may result in disqualification. Late registration or submission may be rejected. Judging decisions are final after eligibility and originality checks. Intellectual property: The prize-winning submission assigns all rights to LV8 Tech when the winner accepts the prize. All submissions grant LV8 Tech a perpetual, irrevocable, royalty-free licence to the concepts and architecture contained in the submission. Participants warrant that their work does not infringe third-party rights. If you do not agree, do not register or submit. Clarifications: Material clarifications will be shared through the common participant channel so every participant receives the same information. Contact: lv8tech@gmail.com