🤓 Latest Submissions
Moyu Pocketful Factory
Pocketful Factory is a fully autonomous zero-human-code software factory built for the WeAreDevelopers x BAND Hackathon, competing in the Pocketful Track. Operating directly inside BAND Desktop, the factory coordinates a three-agent autonomous swarm consisting of Architect, Coder, and Auditor under generic, domain-independent mandates. The human operator dispatches only high-level requirements into the shared room. From that initial prompt, the swarm completely autonomously drives design, implementation, adversarial testing, and verification: 1. Autonomous Swarm Orchestration: Inside the BAND room, the Architect formalizes double-entry balance conservation, bitemporal state snapshots, and idempotency write contracts. The Coder implements the core financial ledger, optimistic balance updates, funds hold/capture, and self-healing error recovery. 2. Genuine Adversarial Code Review: The Auditor executes continuous red-teaming directly in the room. The immutable room log (room.json, containing 464 multi-turn events) proves genuine adversarial rigor: early commits were rejected twice for atomicity and rollback invariant leaks, prompting the Coder to repair the issues before final approval of commit 7b671ec. 3. 100% Conformance & Test Pass Rate: The service achieves a 100% pass rate across all 193 official harness conformance tests spanning all 4 stages, with zero regressions, zero manual workarounds, and zero track leaks. 4. Clean Production Packaging: The solution delivers fully isolated, self-contained Docker container environments capable of serving health probes and the interactive Web UI within seconds without outbound network dependencies. Pocketful Factory demonstrates the future of dark software manufacturing: resilient financial software verified through multi-agent adversarial consensus.
6 Oct 2026
.png&w=828&q=75)
Production RAG Engine on AMD ROCm
A high-performance, production-grade Retrieval-Augmented Generation (RAG) engine architected for the AMD AI Academy Challenge (Mini-Challenge 3), achieving a perfect 200/200 score on real AMD GPU hardware with sub-second retrieval latency and 2.94-second corpus indexing. Key Highlights & Architecture: 1. Multi-Format & Multimodal Ingestion: Safely ingests and structures complex heterogeneous enterprise data including PDF, DOCX, XLSX, CSV, LOG, Python source code, and hardware schematics. Integrates an offline RapidOCR pipeline with 2D spatial coordinate mapping to reliably extract dense pinout assignments and printed revision labels without internet access. 2. Deterministic Retrieval & Multi-Hop Reasoning: Combines BM25 token relevance with deterministic entity extraction and multi-hop causal reasoning to bridge diagnostic error logs directly to RMA parts and resolved bug tickets. 3. Necessity Citation Pruning: Employs strict citation pruning to eliminate retrieval noise and false positives, delivering 100% precision. 4. Robust Security & Lifecycle Isolation: Gracefully identifies and rejects obsolete or superseded documents, while providing disciplined refusal on encrypted documents. 5. Strict Hardware & System Compliance: Built on the mandated ROCm 10.0 Ubuntu 26.04 PyTorch runtime, strictly maintaining between 1.0 and 48.0 GiB of active GPU VRAM.
1 Dec 2026
