How AI Models Learn to Do Real Work

the vibe

AI models are getting smarter. But there’s still a big gap between performing well on a benchmark and reliably doing real-world work. Join us for an evening of networking, drinks, and a fireside conversation with Osvald Nitski, Head of Product at Mercor, and Chris Clark, Co-Founder of OpenRouter, about what it takes to close that gap. Together, they’ll explore how the way we train, evaluate, and deploy AI is changing - from benchmarks and expert data to reinforcement learning environments, model choice, and agents that can take on increasingly complex work. We’ll dig into the big questions shaping what comes next: How should we measure what models can actually do? How do models learn from real-world environments? And what changes when AI can reliably take on work that once required humans? Expect a thoughtful discussion, a great crowd, and plenty of time to connect with founders, researchers, engineers, and technical leaders building the next generation of AI products. *** OpenRouter is the unified interface for LLMs: 500+ models from 80+ providers, processing 400T+ monthly. Better prices, better uptime, no subscriptions.