https://video.ut0pia.org/videos/watch/a13e1eef-7a9e-4810-8135-3d28521a859b To train an agent that can run production software, you need training data that looks like production, and that is what Joseph Wang's team at Emulated builds. Coming from network infrastructure backgrounds, they know what happens when something like a database goes down at scale, and they argue that current post training environments do not capture it. A real task is not a tidy code diff; it is fifty to a hundred turns of solving live traffic while distributed nodes fail, configs conflict, and unforeseen problems appear mid incident. So Emulated simulates whole companies. Imagine acting as an engineer inside a cloud provider or an infrastructure service, provisioning resources across VPCs, subnets, and security groups, meeting real bars around cost and deployment, and keeping a service alive as it grows, all inside a high fidelity environment rather than a stub. Wang's bet is that domain expertise plus faithful simulation is what lets agents learn the messy, end to end reality of infrastructure work, and he closes looking for people who have trained models or run real infrastructure to help push that fidelity further across more domains. Speaker info: https://emulated.so/, Timestamps: 0:00 - Useful work over longer horizons 1:20 - Backgrounds in network infrastructure 2:26 - How environments shape capability 3:16 - Fifty to a hundred turn tasks 4:59 - Why real incidents are messy 7:11 - Real infrastructure isn't a code diff 7:40 - Acting as an engineer inside the cloud 9:37 - Deployment, cost, and scaling bars 13:29 - Why it's called Emulated 15:01 - Simulating full companies Sat, 01 Aug 2026 17:56:20 GMT https://validator.w3.org/feed/docs/rss2.html PeerTube - https://video.ut0pia.org https://video.ut0pia.org/lazy-static/avatars/0287a09a-aae7-4840-9843-b416426e7046.webp https://video.ut0pia.org/videos/watch/a13e1eef-7a9e-4810-8135-3d28521a859b All rights reserved, unless otherwise specified in the terms specified at https://video.ut0pia.org/about and potential licenses granted by each content's rightholder.