@Prashant Suryawanshi and @Yossi Benvenisti maybe it might be interesting you.

@Fred Warsaw Welcome Fred. Networking background is actually a strong foundation for agent work, most production failures are infra problems, not AI problems. Path for you: finish Python on Udemy, then jump into the OpenAI SDK course in the classroom. Vibe Coding Bootcamp ships in 4 weeks and covers script-to-shipped-app. PS: you're locked in lifetime free as a founding member (198/500 then paid). Welcome in

Quick thoughts on AWS Desi Quest: 1. Antigravity vs Cursor: both have strengths. Antigravity is sharper on multi-file context where it reasons about your full project. Cursor is faster for tight single-file iteration. Both miss things, just different things. 2. Google AI Ultra rate limit: mix providers to spread the load. Use Gemini for the cheaper background work (quiz generation, simple lookups), Claude for the reasoning-heavy work (architectural lens, security threat modeling). Reduces the chance of hitting any one quota wall. 3. The UI architecture you described (sessions left, modes right, flashcards in chat) is sound. Mirrors how Linear and Notion structure attention: persistent context on the left, mode-specific actions on the right, conversational generation in the middle. Don't reinvent that pattern, build on it. 4. "notebooklm python version as backend" — curious. Using the unofficial google-genai library to hit NLM endpoints, or building your own retrieval layer that mimics NotebookLM's behavior? The latter is more portable. For collaborators: anyone here building cert-prep tools, study apps, or multi-mode AI interfaces, reply to Aakash. He's actively iterating on something useful. Will feature this in Show Your Build Sunday if you're up for it.