Log in or create a free Rosenverse account to watch this video.
Log in Create free account100s of community videos are available to free members. Conference talks are generally available to Gold members.
This video is featured in the Evals + Claude playlist.
Summary
If you’re a product manager, UX researcher, or any kind of designer involved in creating an AI product or feature, you need to understand evals. And a great way to learn is with a hands-on example. In this talk, Peter Van Dijck of the helpful intelligence company will walk you through writing your first eval. You will learn the basic concepts and the tools, and write an eval together. This talk is hands on; you can follow along, and there will be plenty of time for questions. You will go away with an understanding of the basic building blocks of AI evals, and with the confidence that you know how to write one. And more importantly, you’ll build some intuition, some product sense, around how the best AI products today are built, and how that can help you use them more effectively yourself.
Key Insights
-
•
Evals consist of a task, a golden dataset with known correct outputs, and an evaluator that measures correctness.
-
•
Manual AI prompt testing is slow and inconsistent; automated evals accelerate and scale evaluation.
-
•
UX and product teams can and should learn evals as a practical, non-technical skill.
-
•
Creating your own golden dataset is essential and cannot be outsourced or fully automated.
-
•
Models are fixed once trained; improvements happen by refining prompts and context design, not retraining the model.
-
•
Evaluations measure task performance, not the underlying model itself, allowing comparison across models.
-
•
Outputting a confidence score from models is unreliable due to lack of internal memory and inconsistent scale interpretation.
-
•
Biases are baked into models during training via evals used in post-training refinement.
-
•
LLMs can be used to judge other LLM outputs to evaluate tasks with non-binary answers.
-
•
Effective eval work requires collaboration across data analysts, engineers, subject matter experts, and UX/product teams.
Notable Quotes
"Evals are like a way to define what good looks like."
"The model was baked and once it’s baked, it does not learn again until they bake a new one."
"You need to be looking at the data. Nobody wants to, but that’s core work."
"Without a golden dataset, you have to build the golden dataset yourself."
"We’re not teaching the model anything; we’re improving our prompts and context."
"Confidence scores from the model are not a good idea because the model has no memory."
"Biases are baked in through the evals used during model training and post-training."
"LLMs judging other LLMs might sound crazy, but if you do it right, it works."
"Evals are a product and UX skill; learning them lets you make these systems do what you want."
"There is a large and growing capability overhang in these models we haven’t discovered yet."
Or choose a question:
More Videos
"People can have the same values and ideas but if you don’t tune them properly, you just don’t get what you want."
Lisa WelchmanCleaning Up Our Mess: Digital Governance for Designers
June 14, 2018
"Most chiefs of staff burn out in two years, so it’s important to be clear with yourself and your leader about your goals."
Lisa GirondaOpener: Chief of Staff–An unexpected journey
January 8, 2024
"Nudge to me now is much more about wiggle it and see what happens rather than expecting exact outcomes."
Jen BriselliLearning Is The Engine: Designing & Adapting in a World We Can’t Predict
April 16, 2025
"The digital sector’s compounded impact on carbon footprint exceeds industries like airlines when aggregated across millions."
Ned Gartside Mike Gifford Zoe Lopez-Latorre Tzviya SiegmanNavigating accessibility and climate
April 17, 2024
"Most B2B research is done remotely and online, so we needed a tool that was easy to use without in-person facilitation."
Sean DolanA Practical Look at Creating More Usable Enterprise Customer Journeys
October 31, 2019
"Every piece of research we’ve conducted represents valuable context and insights that can still guide us today."
Katie HansenFinding the unknown in the known: Harnessing meta-analysis and literature review
March 12, 2025
"Small companies like Descartes Labs built supercomputers entirely in AWS with no physical servers, showing a profound shift in how computing power is accessed."
Hugh DubberlyProblems with Problems: Reconsidering the Frame of Designing as Problem-Solving
June 19, 2019
"The standard SUS question about needing a technical person to use the system is frustrating because many assistive tech users are very technical themselves."
Sam ProulxSUS: A System Unusable for Twenty Percent of the Population
September 29, 2021
"You need to find your own mattress—a grounding purpose or phrase—that you can rely on when progress feels hopeless."
Barb SpantonDoing Work That Matters: A Look Beyond The Idealistic Notion of 'Doing Meaningful Work'
June 10, 2022
