<p><em>Jonathan Wall</em> is the CEO at <a href="https://runloop.ai/" target="_blank" rel="ugc noopener noreferrer"><em>Runloop.ai</em></a>, working on enterprise-grade infrastructure and execution environments for AI coding agents.</p><p><br></p><p>The Future of AI Agents is Sandboxed // MLOps Podcast #353 with Jonathan Wall, CEO at Runloop.ai.</p><p><br></p><p>Join the Community: </p><p><a href="https://go.mlops.community/YTJoinIn" target="_blank" rel="ugc noopener noreferrer">https://go.mlops.community/YTJoinIn</a></p><p>Get the newsletter: <a href="https://go.mlops.community/YTNewsletter" target="_blank" rel="ugc noopener noreferrer">https://go.mlops.community/YTNewsletter</a></p><p><br></p><p>Shoutout to <a href="https://studio.youtube.com/channel/UCDanFXxCpNK1p_QHm0i9gzA" target="_blank" rel="ugc noopener noreferrer"> @runloop-ai </a> for powering this MLOps Podcast episode.</p><p><br></p><p>// Abstract</p><p>Everyone’s arguing about agents. Jonathan Wall says the real fight is about sandboxes, isolation, and why most “agent platforms” are doing it wrong.</p><p><br></p><p>// Bio</p><p>Jon was the techlead of Google File System, a founding engineer at Google Wallet, and then the founder of Inde, which was acquired by Stripe. He is building Runloop.ai to bridge the production gap for AI Agents by building a one-stop sandbox infrastructure for building, deploying, and refining agents. </p><p><br></p><p>// Related Links</p><p>Website: <a href="runloop.ai" target="_blank" rel="ugc noopener noreferrer">runloop.ai</a></p><p>Blogs and content at <a href="https://www.runloop.ai/" target="_blank" rel="ugc noopener noreferrer">https://www.runloop.ai/</a></p><p><br></p><p>~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~</p><p>Catch all episodes, blogs, newsletters, and more: <a href="https://go.mlops.community/TYExplore" target="_blank" rel="ugc noopener noreferrer">https://go.mlops.community/TYExplore</a></p><p>Join our Slack community [<a href="https://go.mlops.community/slack" target="_blank" rel="ugc noopener noreferrer">https://go.mlops.community/slack</a>]</p><p>Follow us on X/Twitter [@mlopscommunity](<a href="https://x.com/mlopscommunity" target="_blank" rel="ugc noopener noreferrer">https://x.com/mlopscommunity</a>) or [LinkedIn](<a href="https://go.mlops.community/linkedin" target="_blank" rel="ugc noopener noreferrer">https://go.mlops.community/linkedin</a>)]</p><p> Sign up for the next meetup: [<a href="https://go.mlops.community/register" target="_blank" rel="ugc noopener noreferrer">https://go.mlops.community/register</a>]</p><p>MLOps Swag/Merch: [<a href="https://shop.mlops.community/" target="_blank" rel="ugc noopener noreferrer">https://shop.mlops.community/</a>]</p><p><br></p><p>Connect with Demetrios on LinkedIn: <a href="https://www.linkedin.com/in/dpbrinkm/" target="_blank" rel="ugc noopener noreferrer">/dpbrinkm</a></p><p>Connect with Jon on LinkedIn: <a href="https://www.linkedin.com/in/jonathantwall/" target="_blank" rel="ugc noopener noreferrer">/jonathantwall/</a></p><p><br></p><p>Timestamps:</p><p>[00:00] GitHubification of workflows</p><p>[00:29] Sandbox definitions explained</p><p>[04:47] Agent setup explanation</p><p>[08:03] Sandbox vs API agent</p><p>[13:51] Resource usage in sandbox </p><p>[22:50] Agent evaluation setup</p><p>[28:08] Failure cases value</p><p>[31:06] Sandbox isolation vs multi-tenancy</p><p>[36:14] Frameworks vs Harnesses</p><p>[39:02] Langraph vs Harness comparison</p><p>[43:22] Agent flexibility and verification</p><p>[52:51] Training data focus</p><p>[57:10] Wrap up</p>