Case Studies

Selected work and experiments from Sybil Solutions (by 0xSero).

MiniMax Agent – Agent Communication Protocol Integration

Integration of MiniAgent with Agent Communication Protocol (ACP), enabling seamless agent execution within Zed editor through Anthropic-compatible API.

1,000+

AIDeveloper ToolsTypeScript

Open Queue – Message Queue Plugin for OpenCode

A TypeScript plugin that queues messages while OpenCode is thinking, preventing context confusion from interrupted responses.

AI Coding Assistant Training Data Extraction Toolkit

A Python toolkit for extracting conversation histories, code contexts, and metadata from popular AI coding assistants for ML training and analysis.

AIPythonLLM Infrastructure

MiniMax-M2 Proxy – Bridging 229B Models to Standard APIs

A translation proxy that enables MiniMax-M2 (229B MoE model) to work seamlessly with OpenAI and Anthropic SDKs through intelligent XML-to-JSON conversion.

Open Orchestra – Multi-Agent Orchestration System

A hub-and-spoke multi-agent orchestration plugin with Neo4j-backed memory, 22+ tool APIs, and built-in worker profiles for complex AI workflows.

Azul – AI-Powered Terminal Web Browser

A terminal-based web browser written in Rust with integrated AI chat, tool-calling, and multi-engine search for hands-free browsing.

OpenAI, Ollama, Custom

AI Providers

AILLMModel Compression

Compressing GLM & MiniMax to Run on a MacBook

End-to-end documentation of compressing GLM-4.7 and MiniMax from 717GB BF16 to ~92GB using REAP expert pruning and AutoRound INT4 quantization, achieving 7× reduction while maintaining model quality.