Hosted dinner | BY INVITATION | NEW YORK

The agent worked in the demo and not on Monday

Name: The agent worked in the demo and not on Monday
Start: 2026-09-24T18:30:00Z
Location: New York

Heads of AI and applied research compare what actually broke when they moved an agent from a controlled evaluation into production, and what they now catch before ship.

Date: Sep 24, 2026
Time: 18:30 to 22:00
Location: New York
By invitation only: 14 spots, 4 open

Request an invitation

Agenda

What did your evals miss that a customer found in week one?
Where did retrieval or tool use fail in a way you did not see coming?
Which incident cost you the most trust internally, and what did you change the day after?
What is the next class of failure you are already writing an eval for?

Who will be there

12 to 14 heads of AI and applied research leads from labs and applied teams building on frontier models.

Head of AI or machine learning

Research lead

Engineering leadership

Heads of AI and applied research compare what actually broke when they moved an agent from a controlled evaluation into production, and what they now catch before ship.

Further details

Chaired by Meridian's head of research. Peer-led, no vendor presentation. No demo, no roadmap slide, no follow-up sequence.

Restaurant in Tribeca, confirmed on the day. Doors 18:30, seated dinner 19:00, close 22:00.

Dinner, wine and travel within Manhattan are covered by the host.

About Meridian Labs

Meridian Labs builds evaluation and observability tooling for teams shipping agents into production. We host these rooms because the questions that matter are the ones nobody wants to put on a slide.