Positive AI Labs Workshop: Building AI Evaluations for Human Flourishing
Start time: Saturday Jan 24th, 2026 @ 9:30am
Location: San Francisco
Why We're Coming Together
Together, we build tests to guide frontier AI models on human flourishing that will touch 5+ billion humans. Please join us to ensure the world’s most powerful AI models are shaped by the best of human values. Three important groups will come together for this effort:
Flourishing Stewards (e.g. philosophers, ethicists, contemplative tradition teachers, neuroscientists, economists)
Domain Stewards (e.g. mental health, conflict mediation, education, health advice)
Evaluation Experts from AI labs
The Goal
Evaluations are tests that determine what AI models become good at doing. We're building evaluations that can steer frontier LLMs, e.g. ChatGPT, Claude, Gemini, or Meta AI, toward human flourishing in many fields, including teen and adult mental health, purpose-making, re-skilling, conflict resolution, ethical reasoning and more.
By the end of the day, we'll have edited an MVP evaluation blueprint ready for a team to build.
How It Works
Morning: Learn how AI evaluations work, align on success criteria, map flourishing taxonomy to real-world use-cases, draft exemplar tasks and rubrics in breakout sprints.
Afternoon: Design the judging stack and metrics, red-team the spec for exploits, lock the MVP, build the 9-month plan with owners and milestones, define governance and lab adoption strategy.
Who Should Come
We're looking for people who want to roll up their sleeves and develop a future in which AI systems act with wisdom, compassion, and ethical clarity. This only works if a plurality of voices and perspectives are in the room (flourishing stewards, domain stewards, and evaluation experts).
This is a full-day working session - come help us build the future!
About Positive AI Labs
A non-profit building evaluation infrastructure that steers frontier AI toward human flourishing. Our goal is to influence how every major AI model is trained via tests that define what "good" means for complex human situations, when often there is no single "right" answer.
Lunch and refreshments provided.
Questions? Contact:
Theo: [email protected]
Note: Limited capacity. Approval required to ensure the right mix of technical, wisdom, and domain expertise.