OpenAI
AnthropicVerified interview challenge
Design a prompt playground platform
Design a platform where users test prompts across models, compare outputs and run larger experiments.
Unlock with Full AccessThe challenge
Users edit a prompt, compare streamed responses from several models and save runs for later. The same platform must also support experiments with thousands of combinations. Design how those workloads share provider capacity without a large experiment making the interactive editor unusable.
A result needs to remain understandable after the prompt, model or settings have changed. Your design must connect that history to execution, cancellation and provider limits, so later comparisons refer to the work that actually ran.
What you'll practise
Interactive and batch work
Balance quick prompt iterations against experiments that occupy provider capacity for longer.
Run history
Keep the relationship between inputs, settings and results clear as users iterate.
Execution controls
Explain how cancellation, failures and cost limits affect work already in progress.