The Chain of Debate: A Prototype for Autonomous AI Agents
A cornerstone of the Kognys platform is our unique "Chain of Debate" methodology. This approach demonstrates how our AI agents can autonomously manage data, share knowledge, and evolve based on interactions, effectively serving as a working prototype for the challenge. Our AI agents embody the principle that intellectual progress is forged through discussion and critique, transforming peer-review into a real-time, collaborative, and transparent engine for discovery, all powered by the Unibase stack.
Here's how it works, showcasing the challenge criteria in action:
Initial Prompt & Hypothesis (Autonomous Data Management): A researcher provides an initial research question. A primary "Proposer" agent is tasked with autonomously scouring the OpenAlex database to formulate an initial thesis.
Assignment of Adversarial Roles (Evolving Based on Interactions): Using LangGraph, we intelligently assign a team of specialized AI agents to challenge and refine the initial draft, each with a specific persona (The Empiricist, The Contrarian, The Synthesizer). This demonstrates how agents can evolve their roles and strategies based on the research context.
The On-Chain Debate (Sovereign Memory & Interoperability): The agents engage in a structured, multi-turn conversation, orchestrated via the Unibase Agent Interaction Protocol (AIP). This entire debate—every argument, citation, and revision—is immutably recorded in the agents' persistent, sovereign memory using Unibase Membase for complete transparency on the BNB testnet. This directly showcases decentralized memory and cross-platform interoperability within the agent system.
Iterative Refinement (Autonomous Evolution): The Proposer agent (or the Synthesizer) continually revises the paper based on the rich feedback from the debate. This refinement loop continues until the agents reach a predefined quality threshold, demonstrating autonomous evolution.
Final Output with Verifiable Provenance (Immortal & Transparent): The researcher receives a highly-refined paper. Crucially, this final document is stored permanently and verifiably on Unibase DA. The researcher also receives access to the complete on-chain debate history from Unibase Membase, allowing them to understand how the conclusions were reached, evaluate the counter-arguments, and trust in a process that is rigorously vetted and fully transparent. This validates the concept of an "immortal" AI agent and its contributions.
Kognys is actively building the future of scientific research, powered by immortal AI agents with decentralized memory and cross-platform interoperability on BNBChain. We are confident our working prototype and innovative approach directly align with the spirit and technical requirements of this challenge.
Last updated