Voices
Hamel Husain: ''It's Hard to Eval' Is a Product Smell'
Husain argues that when a product is genuinely hard to evaluate, that difficulty is usually a symptom of underlying design problems that also degrade the user experience — reframing evals from a QA afterthought into a product-design signal. Coming from the practitioner who has taught evals to thousands of engineers and PMs, it inverts the common complaint: if you can't eval it, fix the product, not just the harness. Practical for anyone building AI features — treat 'we can't measure this' as a design bug to resolve upstream.
↳ Follow the thread