AAIF Community

Everyone uses AI agents. But where are the companies building their own? Erica Hughberg and @Dpbrinkm discuss vendor agents, MCP vs. A2A, context retrieval, and why coding agents can miss system-level security and stability tradeoffs.

Sam Herring used Hermes Agent to search government auctions until it found him a car. He also explains how agent failures can become RL training environments.

Why the excitement around JEV? It's not that classification is new. It's that developers can access it without building the surrounding infrastructure. Erica Hughberg and @Dpbrinkm discuss skill selection, agent routing, and more.

Cut AI costs by 10x, then watch new features eat into the savings. Ioannis Zempekakis of @Prosusgroup on context management, model routing, and task-level evals.

An AI agent using your logged-in browser can inherit your access to email, banking, and other services. Erica Hughberg and @Dpbrinkm discuss workload identity, scoped permissions, prompt injection, and why agent authorization needs to account for intent.

@mihail_eric on agent-ready MCP design: tools should reflect agent intent, not simply mirror API endpoints.

One user request can trigger dozens of API calls. Agent traffic is bursty, unpredictable, and challenging to price. Erica Hughberg and @Dpbrinkm discuss API gateways, Kubernetes limits, and Google's Agent Substrate.

AI coding agents can diagnose bugs without producing maintainable code. Dexter Horthy of HumanLayer discusses software factories, background-agent infrastructure, and the operational work behind running agents at scale.

AI infrastructure is hitting physical limits: power, cooling, rack density, and hardware supply chains. Zachary Smith of Datum discusses the constraints and how data centers could reuse waste heat.