Evals Are the New Unit Tests: Notes From Snowflake
Notes from the teams shipping agent mode to real users.
The interesting question is not whether Vercel v0 works. It does. The interesting question is what teams do with it once the novelty wears off.
The skeptical read is that we are watching a feature, not a platform. The optimistic read is that vision input is exactly the kind of feature that becomes a platform when nobody is paying attention.
Booking.com has been quietly running data cleanup through Warp for months. The results are unglamorous and, for that reason, more interesting than another benchmark chart.
Eval harnesses, once an afterthought, are becoming the most important piece of code in many AI projects. Shopify's team treats theirs the way an SRE team treats a runbook.
Inside Figma, the rollout looked less like a moonshot and more like a slow migration. A pilot, a champion, a quiet expansion, a budget line.
The cost curve matters here. Grok 4 is roughly an order of magnitude cheaper per token than the equivalent model 18 months ago, and that changes which problems are worth automating at all.
Eval harnesses, once an afterthought, are becoming the most important piece of code in many AI projects. Intercom's team treats theirs the way an SRE team treats a runbook.
None of this guarantees a clean story. Alibaba Qwen could ship a model next month that rearranges the assumptions in this piece. But the direction of travel, for now, is clear enough to plan around.