ZenML Opens the Coffee Table to the Agent Failures Teams Stopped Trusting
An agent failure can become a ticket, a trace, a screenshot, or a story the team repeats until everyone remembers the conclusion differently. The difficult work begins when somebody has to turn that production moment into a test the next version cannot quietly fail again.
ZenML founders Hamza Tahir and Adam Probst are hosting a small Friday coffee for practitioners to compare how they evaluate models and agents, what is breaking, and which signals they have stopped trusting. There are no slides and no pitch deck. The subject is the loop between what happened in production and what the team changes before the next release.
ZenML describes Kitaru as replay-based evaluation software that imports production traces and turns observed failures into evaluators. The team is also building Kaizen, an autonomous AI engineer intended to watch agent traces, improve evaluations, and fine-tune open-weight models with people remaining inside the loop. The coffee gives those ideas a harder surface: practitioners carrying real failures, uneven evidence, and systems that rarely improve as cleanly as the product diagram suggests.
A better agent may begin with a better model. It may begin earlier, when the team learns how to preserve the failure clearly enough that the system cannot make everybody discover it again from scratch.
Event details
- Date: Friday, October 9, 2026
- Time: 11:00 PT
- Location: Mission, San Francisco Bay Area, CA
- Format: Breakfast, Brunch or Lunch, Networking, Panel / Fireside Chat
- Host: ZenML GmbH
- Registration: Official SF Tech Week event page
Registration status, capacity, meeting details, and program information can change. Use the official event page for the latest organizer updates.
Frequently Asked Questions
Who is hosting Coffee and Evals?
ZenML GmbH is the listed host, with founders Hamza Tahir and Adam Probst named in the description.
Will there be presentations?
No. The organizer describes a small discussion with no slides and no pitch deck.
What is the conversation about?
The group will compare model and agent evaluation practices, production failures, and signals teams no longer trust.
Where the Money Moved
The intelligence briefing of the innovation economy. Funding, M&A, debt and fund closes, read as market signal rather than deal announcements.
Subscribe to Where the Money Moved







