by matiasmolinas on 6/25/25, 6:52 PM with 1 comments
by matiasmolinas on 6/25/25, 6:52 PM
Current experiments:
• LLMunix - Pure markdown operating system where Claude interprets markdown documents as functional OS components. Everything is either an agent or tool defined in markdown specs.
• EAX Router - Intelligent LLM routing that automatically selects optimal models based on task requirements. Implements the "right model for the right job" principle with cost/latency/quality optimization.
• EAX Marketplace - Decentralized auction protocol where AI agents bid on tasks competitively. Creates natural specialization and discovery mechanisms - the "AI council meetings" Suleyman envisions.
• SAL-CP - Self-aware communication protocol enabling rich context sharing between agents. Agents communicate not just what they're doing, but how they're thinking and what they need from collaborators.
• Framework Core - Foundational tools for building adaptive agent systems with memory-driven learning and behavioral constraints.
• Agent Examples - Practical implementations showcasing multi-agent coordination patterns and real-world applications.
Research focus: Moving from hardcoded agent chains to dynamic, market-based coordination. Our hypothesis: competitive and collaborative agent ecosystems will be more resilient, efficient, and adaptive than current rigid architectures.
All experiments are permanently alpha status - we're focused on exploring concepts and generating research insights rather than production deployment. Apache 2.0 licensed.
Looking for researchers interested in multi-agent coordination, LLM routing strategies, auction mechanisms, or markdown-based system architectures.
What coordination patterns do you think will emerge as AI becomes truly collaborative?