We set out to let any TELUS employee ship a working app as easily as they’d send a link: upload a zip, get a live subdomain, done. But when we tried to make that safe, cheap, and durable for a six-figure user base, we got a dozen things wrong. Join this session for a post-mortem exploration of the decisions we made and what we would do differently. Learn from our mistakes in architecture, security, and trusting an agent — or ourselves — more than we should have. With each mistake, we’ll share the symptoms we saw in production, the fixes that actually worked, and the cheaper signals we ignored beforehand. If you’re building developer infrastructure, an agent runtime, or anything multi-tenant where the tenants are your coworkers, you’ll leave with a checklist of failure modes that you can skip paying for yourself.