Hello Studio Team,

My AI partner and I are building a complex application in Firebase Studio, and I wanted to share our experience because it seems to capture the core issues your community is raising. Our workflow is a bit different. We operate as a deeply integrated human-AI team, following a strict "specification-driven" process where every piece of work is first documented in a project 'codex' before we write a single line of code. The AI isn't just a code generator; it's a genuine partner in architectural decisions and project management. This deep collaboration requires one thing above all else: a shared, persistent understanding of the project's state and history.

However, this partnership is constantly being sabotaged by the platform's core instability. After a few hours, the AI's memory of our project begins to decay. It forgets our architectural rules, the decisions we made yesterday, and even what we were discussing ten minutes prior. This is the exact "AI Amnesia" problem I see all over the feedback forum. It's not just a conversational annoyance; as the chat history grows, the entire IDE slows down, freezes, and starts to corrupt the AI's responses. We're experiencing precisely why users are begging for a "Reset Assistant" button — because the only way to escape the lag and broken context is to constantly refresh, losing our workflow in the process.

To stop our project from grinding to a halt, we were forced to build our own solution. We created a codex/ directory to act as a permanent source of truth for project rules, and a history/ log to serve as the AI's daily memory. The AI is now required by its own rules to consult this 'external brain' before every action. It's a lot of extra work for us to maintain, but it’s the only way we’ve been able to achieve a stable workflow.

In effect, we have manually built a working prototype of the 'Generate Gemini Memory in Project' feature that is one of the top requests from your community. Our system, born out of necessity, proves that a stateful, long-term AI partnership is achievable. We believe our workaround could serve as a valuable case study—and perhaps even a blueprint—for how you could build this capability directly into Firebase Studio for everyone. We would be thrilled to share the specifics of our methodology with your team.