Wayfinder: A new AI Agent Network

Key Insights

• Parallel released a new agent network whitepaper that describes the Wayfinder network
• Wayfinder works by incentivizing developers to build micro blocks of code that interact across various blockchain networks and are subsequently usable by Wayfinder AI agents
• Execution paths are economically secured by the Wayfinder native token and further token demand is driven by its use to purchase new agents (deemed shells) and its use as a gas token
• The overall Wayfinder model is interesting as it presents a new way to package and sell software broadly and does so in a way that accrues network effects and value to the network

Within crypto, AI has been a hot sector, growing from a $4.6 billion total circulating market cap to over $26 billion in just 9 months. Most of the hype and liquid market valuation have been concentrated on infrastructure plays such as underlying GPU networks (Render, Akash, and Nosana) or inference networks that either host or incentivize the execution of models (Bittensor, Ritual, etc).

What’s largely been lacking is applications that effectively productize model inference(s) into useful consumer products. However, I’d argue that a single LLM inference, or even a few series of LLM inferences, isn’t all that valuable as an inference simply condenses or transfigures text into another form. What is valuable is many, many inferences each with specially tuned domain-context or specific logic chained together. Especially when executed in a non-deterministic way.