An evening with Auth0, MongoDB, Vellum, and founders building agent-based products put on by AI Tinkerers.

There’s a story circulating right now about what LLMs can do at scale - finding zero-days, operating autonomously, acting faster than any human security layer can follow. Some of it is hype. Some of it isn’t. And the teams building agent-based products are the ones who need to think most clearly about which is which.

We;re bringing together 15 to 20 early-stage founders, CTOs, and technical leads for a private chef omakase dinner in a downtown artist studio - built around that question, and everything it pulls into focus.

The Premise

The Claude mythos narrative has a moment right now. But what’s easy to miss is that the real story isn’t just about model capability - it’s about scale, autonomy, and all the decisions that happen when an agent is acting without a human in the loop.

That’s exactly where identity becomes load-bearing. If an agent holds credentials, invokes tools, delegates to sub-agents, and takes actions across systems - who authorized it? What policies govern it? And when something goes wrong, is there an unambiguous audit trail for what happened and why?

Auth0 has been deep in this problem. Their work on agent-as-principal maps directly onto what a lot of teams are hitting in production right now. But identity is only half of it. An agent acting over time needs somewhere to hold its memory, its state, and the record of what it did - and that’s the data layer. MongoDB has been building for exactly this: a flexible document model and vector search that let agents store, retrieve, and reason over their own context, which is also where that audit trail actually has to live. And Vellum sits where all of this gets built - the workflows, evals, and orchestration that turn an agent from a demo into something you can actually ship and trust in production.

We figured the right place to work through it was a long table, a private chef, and a room full of people who are actually building these systems.

What to Expect

AI Tinkerers VIP Dinners are intimate, curated gatherings. Each one is built for signal over scale.

This means:

A private seated dinner with open, peer-level conversation

A few light framing prompts, no talks or slides

Builders trading notes on what they’re building, what feels fragile, and what they’re rethinking

Conversations that continue long after the plates are cleared!

Who’s At The Table

Early-stage founders, CTOs, and technical leads actively building agent-based products. The room is for people shipping, making architecture decisions, and thinking seriously about how these systems should behave in the real world.

Event Details

Date:June 4th 2026

Time: 6-9pm

Location: New York City (Exact location shared upon approval)

Capacity: Limited to 20 attendees

This is a seated, curated dinner. To maintain a strong room, every attendee is screened to ensure the table is made up of active builders and technical leaders.