​Agents Done, Accuracy None: Can Evals Make AI Agents Reliable?
Frontier Tech Series · Session 01

​📅 Tuesday, November 17 · 5:00 – 7:00 PM
📍 Venture Dock, 555 University Avenue, Palo Alto

​About the Event

​AI agents sound confident. They also get things wrong all the time. As teams move agents from demos into real products, the gap between “it works in the demo” and “it works every time” is where most of the pain lives.

​So here’s the question: can evals actually make agents reliable, or are we measuring the wrong things?

​A new series for our community

​This is the first session of the Frontier Tech Series at Venture Dock, a new format we’re starting for our community. Each session is a small-group conversation on one sharp question about emerging technology. It’s an intimate table of 4 to 5 people, 10 at most. There are no talks, no panels and no audience. Everyone at the table is a participant.

​Who should join

​We’re looking for people building, evaluating or deploying agents: founders shipping agent products, engineers writing evals, researchers working on reliability, and anyone who has watched an agent fail in production and has a view on why.

​Because seats are so limited, we review every registration to make sure the table has the right mix of perspectives.

​Ground rules

​No pitches or sales conversations. Questions over presentations. Ideas from the room can be shared afterward, but nothing is attributed to anyone without their consent.

​Moderated by Sanchit Garg

​Sanchit is the CEO of Zime, where he builds AI for go-to-market teams. He works hands-on with how AI tools and agents perform in real workflows, and recently led an AI Tools for GTM workshop for founders at Venture Dock. He’ll guide the conversation, keep it going deep, and make sure every voice at the table is heard.

​What is Venture Dock

​Venture Dock is a founder-first community and fund, based in Palo Alto with a second location in Milpitas. We bring founders, operators and investors into the same room on a regular basis, through weekly Build From The Dock sessions, 1:1 Founders Lab time with experts, and direct access to our venture partners.

​The Frontier Tech Series is a new part of that. Where Build From The Dock brings 60 to 80 founders together each week, these sessions go the other way: a handful of people, one question, and enough time to actually work through it.

​Become a member

​If you’re building something and want to be around people who are doing the same, we’d love to have you.
👉 Apply for Venture Dock membership