Complete API setup, database setup with MCP, tests, documentation in one prompt! Feels way more agentic, performs way more thinking, feels like automation level 2 unlocked.
3 Likes
Right, yes its level 3 automation 
and the state of the art coder 
2 Likes
That’s impressive to hear. Curious, how consistent is it across different stacks or project sizes? Does it hold up well with more complex, multi-service apps?
Too early to tell as this depends on more than just a project size but also the project structure, framework and programming language you use.
Some report here in the forum great success with Claude 4 Sonnet, and others have trouble doing even small tasks. That can be all depending on prompt, etc.
Its important to read the anthropic guide how to prompt CLaude 4 Sonnet effectively. Trying to use it as other models will likely fail.
It will attempt to create batch processing scripts and modify all code.
Fortunately, when I saw it creating the script, I quickly git committed a copy and restored the code from git before it was modified