Testing with version 3.7 of Claude revealed significant improvements in its gameplay, making it much more capable than previous versions.

When we started using Claude Sonnet 3.5, we were impressed by its capabilities.

When comparing Claude's performance to random button presses, Claude was only slightly better, indicating it had room for improvement.

With the introduction of Claude 4 models, Claude Code has become better at following instructions and executing tasks accurately on the first attempt, enhancing the coding experience.

3.5 Sonnet was the first model where we dedicated significant effort to improve coding capabilities, focusing on longer horizon coding tasks.

Recent evaluations of Claude 3.5 Sonnet included human evaluations where participants had to make business decisions based on extensive documentation, testing how an AI assistant could influence their choices.

Claude lacks an innate sense of time, so I provide it with information about how long it's been trying to perform a task, which helps it recognize when to reconsider its approach.

The completed line Claude wrote is, "His hunger was like a starving rabbit."

The naming of Claude's new model reflects a desire to distinguish it from previous iterations and avoid confusion with past failures.