Recently, I’ve increasingly come to believe that intelligence is no longer AI’s bottleneck. The systems we build around it are. Input Paradox (1) The first issue is the input paradox. When interacting with AI, if the prompt is highly detailed, the model tends to overfit to the user’s framing and assumptions. If it is too concise, the model lacks the context needed to generate something truly...
The one tool I think does this really well is Google’s NotebookLM. But what if I want a single workspace where I can both write and grow my files over time, instead of using NotebookLM more like a study tool for exams?
In IDEs, people use semantic embeddings, .md files and all sorts of tricks to keep the AI low-cost while understanding the whole codebase. So I made a writing workspace inspired by NotebookLM and Cursor. I can keep adding documentation while writing, and the AI uses semantic search in the background to keep costs low while still staying in context. (lemona.studio itself is free, you just use your own API key. For me, after a month, it’s been around $10 while working with thousands of pages of documentation using Gemini 3 Flash.)