Join us for another awesome Agents and APIs Developer Meetup brought to you by Postman! 👩🏽🚀
You'll have a great time learning, connecting with our special guests, making new friends, watching live demos, and enjoying delicious food!
Speakers:
👉 Ben Backus, Software Engineer, Greptile
📝 Remembering Is Easy. Forgetting Is Hard.
At Greptile, we're building an essential tool for vibe-coding production grade code. To stay competitive in this environment, you have to ship faster than ever, but you can't ship slop. While fighting to stay at the bleeding-edge of developer tools, we've learned some hard lessons about how LLMs write code. Conflicting assumptions. Partial updates. Ignored constraints. These are bugs that don’t come from bad code, but from bad memory.
This talk argues that many real-world failures in AI coding systems reduce to a missing primitive: memory that can handle contradiction and change. Ben will give a high level explanation of how Greptile works, walk through the war stories, the workarounds we’ve developed, and a proposal for memory systems that decide what to forget, not just what to remember.
👉 Ben Warren, Founder, Mesa
📝 Filesystems for Agents
The best AI agents today run in sandboxed environments and use flexible, general purpose tools like bash, code, and the filesystem. But when you have lots of ephemeral agents, how do you give them shared, secure access to a filesystem, version work, and keep your infrastructure performant? Join us as we talk about how Mesa designed a virtual filesystem optimized for agents, and our lessons from partnering with top agent builders & sandbox providers.
👉 Nihar Kumara, Sr. Software Engineer, Postman
📝 Building with AI in Postman's native git integration
When you want an AI agent to edit something in your app, the typical approach is to build a bunch of custom tools — one for each action the agent might take. That works, but it adds up fast: more tools, bigger prompts, more tokens, more things to maintain. At Postman, we took a different approach. Instead of teaching the agent how our app works, we gave it something it already knows how to use and wired the app to react to changes. I'll walk through how this plays out in practice with Postman's native git integration, and why the best collaboration happens when neither side has to think about it.
👉 Sohil Kshirsagar, Co-founder & CTO at Tusk
📝 Production Traffic as Your Test Suite
Coding agents can write tests, but they're working from local context and trivial happy path examples, not the behavior your API sees in production. Tests pass, but things still break. Tusk Drift records real HTTP requests, database queries, and cache calls from production, then replays them deterministically against your PR in CI in under 50ms per test, no live dependencies required. When a response drifts from what production expects, AI classifies whether it's an intended change or a regression you're about to ship. We'll walk through how replay-based testing works under the hood and what we've learned about closing the gap between "tests pass" and "nothing breaks in production."
Space is limited. Register now, and we will see you there!
💬 Keep the Conversation Going
Join us in the Postman Discord server to gain access to resources shared by the speakers.
👩🏽⚖️ Legal Stuff
