Fulcrum is an agentic debugger for AI systems, building red-teaming agents that run experiments to diagnose why AI agents fail. The platform plugs into environments, agent traces, and agent source code to find root causes, uncover environment bugs, detect reward hacking, and turn investigations into explorable diagnostic reports. The founders met at MIT doing LLM research and developed methods for scalably generating software environments of arbitrary difficulty. Fulcrum was co-founded by Kaivalya Hariharan and Uzay Girit, both MIT graduates, and went through Y Combinator's Summer 2025 batch.