Anthropic’s Playground vs. OpenAI’s: The week-old tool beat the six-year incumbent
On August 18, Anthropic replaced its Workbench, the prompt-testing tool in its developer Console, with Playground. Anthropic did more than The post Anthropic’s Playground vs. OpenAI’s: The week-old tool beat the six-year incumbent appeared first on The New Stack .
On August 18, Anthropic replaced its prompt-testing tool, Workbench, with Playground. Unlike the former, Playground is stateless and forgets nothing, even removing features such as saved prompts, version history, evals, and team sharing. Anthropic allows users until September 1 to export their old Workbench data. OpenAI's Playground, meanwhile, has been around since 2020, and the company announced it would shut down saved prompts and evals on November 30.
Both companies concluded that prompts belong in code, not a web console. The author tested the new Playground against OpenAI's long-standing Playground by building a PR review bot. Anthropic's Playground successfully executed the test, creating Python code that ran on the author's laptop without any edits. OpenAI's Playground, on the other hand, exported the transcript of the session, including prompts, answers, and reasoning, making the generated code ineffective.
Anthropic's Playground proved more efficient and reliable, checking a box in the company's "always be launching" goals.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
