Evals Course: Playgrounds vs experiments
In Module five of Braintrust's Evals course, we cover the difference between Playgrounds and Experiments.
Playgrounds are scratch pads for exploration. They're great for testing prompts and scoring live, but nothing is saved when you close the tab. Experiments are permanent snapshots that capture the exact prompt, outputs, and scores the moment you hit save.
We'll discuss why this distinction matters for comparing approaches, tracking iteration history, and avoiding costly eval re-runs as your dataset and scoring complexity grow.
Timestamps:
0:00 — Intro: Playgrounds vs. Experiments — what's the difference?
0:07 — What the Playground is: exploring prompts and scoring live
0:21 — The problem with Playgrounds: nothing is saved if you close the tab
0:31 — Think of Playgrounds as a scratch pad for evals
0:35 — How Experiments work: permanent snapshots of config + results
0:45 — Why the comparison view worked without re-running anything
1:00 — Reason 1 experiments are valuable: comparing approaches side by side
1:13 — Reason 2: tracking iteration history across versions
1:18 — Reason 3: avoiding expensive eval re-runs (10–30 min, real API costs)
1:34 — Recap: Playgrounds = explore, Experiments = save & compare
1:42 — What's next: recreating the same eval in code using the Braintrust SDK