I’ve been using Anti-Gravity successfully over the last week.
Last night I upgraded to Business Ultra.
My method and workflow have not changed. In fact, I plan more now and research more brownfield solutions before making changes. I create self-contained implementation plans and then use workflows to validate solutions and mitigate the risk of accidental breakage.
For whatever reason, the LLM has become extremely unreliable. It’s breaking my app all over, admitting that it’s not strictly following its own implementation plans, struggling with regression, and confessing that its biases are the reason for all of its mistakes!
@Google Dev Team: Is this intentional?
Any help you can offer would be great. I’ll give it a week; I have no choice but to request a full refund, so that I can jump back to a more reliable competitor.
Codex has also deteriorated since 5.2. Anthropic is still doing a good job.