It's the first time in the history of the world.
Anthropic, finally on the move.
Just recently, A/ finally cut down the Workbench, a developer tool used for almost two years, and renamed it Playground.
On the surface, it's just a name change.
But the real change is that Anthropic is transforming the original, partially engineered Claude API into a visual product that can be directly handled by ordinary people.
Don't write codeYou can also run Claude's API.In the past, Anthropic's "Workbench" had a distinctly high-cold printing.
For the vast majority of non-senior algorithm engineers who want to explore Claude's API capabilities, there is often a long history of engineering difficulties:
Configuring the Python virtual environment, importing the SDK, configuring the API Key, handwriting the JSON Schema, capturing the error reporting... we haven't started verifying the business idea yet, and it's been half a day on the wrong line.
The newly launched Playground is based directly on the open Messages API, and its core logic is four words: open box.
Officials even give a very clear definition: "Try the model or new API functionality before writing any code".
In the new version of Playground, complex underlying capabilities are broken down into intuitive graphical modules:
- Core parameters: Modeling switching (Claude 3.5 Sonnet, Haiku, Opus), Temperature, Max Tokens, System Prompt, all responding to the secondary level via the right-hand slider and input box.
- High-level features include: Tool Use, Tool Calling, Structured Outputs, Structured Outputs, Code Execution, and Web Search templates, which can be tested by drawing and mapping protocols without the need for handwriting.
- Full transparency of the black box: the sidebar supports the ability to view the original HTTP Request and Response newsletters at any time, token consumption, and context call logic.
A more critical killer is that everything you navigate through a web page supports one-click, seamless code exporting into Python/TypeScript/cURL, etc.
The prototyping process for the entire AI application was completely rewritten as:
"Go ahead and get the code".
The high cost of trial and error that has occurred in IDE in the past has been completely flattened into browsers. AI companies are refining the API interfaces that were originally intended for programmers, refining the interactive products available to adults.
The name of the workbenchTo die innocently
Why did Anthropic abandon the Workbench?
Because the three words "workstation" have a discouraging attribute.
The first thing you need to do is learn how to use the tools before you can start working. For those who have just come into contact with a large model API, the name light is half discouraging.
"Playground" is a playground. The subtitle is too clear: come in and play is over.
This is not what Anthropic is doing. OpenAI has a Playground, Google has an AI Studio. Three giants, three different model systems, finally collided on the product name.
The name synchronization is behind the fact that the industry is building the same paradigm:
Chat is responsible for letting ordinary people experience AI. Playground is responsible for turning experienced people into developers.
Anthropic, more sophisticated than OpenAIPlayground is a PR review bot, and the results are interesting.
The Anthropic side: 1.9 seconds to return the correct JSON, with 102 tokens, costing $0.0029, and the exported script can run directly.
OpenAI responded in 6.2 seconds, consuming about 2,900 tokens.
The exported file is filled with the full dialogue record, and the previous content must be manually deleted and printed statements added to run.
More importantly, Anthropic's output is self-contained script. Check into repo, call directly into CI/CD. OpenAI's output is an interaction record, which you have to disassemble yourself.
Anthropic is telling you: Prompt belongs in code, not in a web console.
AI's next warOn the Control Panel
For the past two years, everyone has been chatting at the entrance. ChatGPT, Claude, Gemini, bloody heads.
But the next big thing is who can get the most people to really integrate models into their products.
Anthropic has been busy this year: the Claude Agent SDK enables developers to build Agents using Python and TypeScript, the Messages API adds on-demand compaction mechanisms, automatically compresses long conversations without manual context management.
Modeling capabilities are increasingly complex, and there are more and more API parameters.
This is the meaning of Playground. The bottom is more complex, and users see only a box, a few sliders, a running button.
Looking ahead, when the model calls, the tool calls, the structured output, the agent sorting can all be done through the visualization interface.
The wall between "API users" and "ordinary users" will be thin and almost non-existent.
Before you could develop an AI application, you had to learn programming.
Now all you have to do is click on it and get the code.
RoboInd editorial channelResearch, robotics companies and market intelligence for cross-border sourcing.Explore the channel ↗
Share this
End of editorialExplore more from Global Robotics News ↗