projects. My history with this IDE hasn't been stellar, but a new model often signals a fresh start. Unfortunately, the gap between Google's marketing and the actual developer experience remains cavernous.
Performance and Skill Discovery
Testing began with a simple CRUD generation task. While
finishes in three. Interestingly, the model showed flashes of brilliance by discovering and activating specific skills like pest testing after I prompted it for its omission. However, the fact that it didn't proactively generate tests—a standard for models like
continues to suffer from random crashes and "terminated due to error" messages that force repeated restarts. Furthermore, the quota system is frustratingly opaque. I hit a usage wall after just nine minutes of work on a free plan, with no clear indication of how many tokens or requests I had left. This lack of transparency makes it difficult to justify a move to their paid tiers.
, Google's progress feels sluggish. The IDE lacks robust support for custom Model Context Protocol (MCP) servers, limiting its extensibility for specialized frameworks. For now,
