aadityaubhat

Oh thanks for the headsup.

No it does not share the underlying data with the LLM. It does share intermediate outputs with the LLM.

That's a good question. Cursor doesn't work with notebooks, it can't create, update or execute notebooks, Bloom supports notebooks natively. Cursor is created for software engineers, while Bloom is created for Data Scientists and Analysts.

I was tired of using ChatGPT/Cursor to generate code and then copy-pasting it into notebooks to actually run things, so I built Bloom.

Bloom is a local IDE for data science with an AI agent built-in. It works with Jupyter notebooks, helps you write SQL/Python, runs code, plots, and can even generate reports and dashboards.

The macOS installer is live on the website, the beta is free for a limited time. Would love feedback on what’s useful, what’s broken, and what you’d want to see next.

TLDR:

- No significant improvement in code generation over Cursor.

- Cursor outperformed Claude on pure backend changes.

- Claude excelled in handling integrated backend and frontend changes.

- Both tools performed well on pure frontend changes.

Demo video - https://www.youtube.com/watch?v=XiomNY1v2J0

https://arxiv.org/abs/2406.02061v1 research paper shows the reasoning breakdown in SOTA LLMs by asking a simple question, “Alice has N brothers and she also has M sisters. How many sisters does Alice’s brother have?” I investigated performance of different prompts on this question, and show that 'Expand-then solve' prompt significantly outperforms standard and chain-of-thought prompts.

Good explanation I found on YouTube few years back - https://youtu.be/0Uqf71muwWc?si=eLDi_zvMtXJeh_WZ

Auto-Analyst is a powerful UI tool that simplifies the process of running analytical queries on your data using plain English. With Auto-Analyst, you can easily perform complex data aggregations, visualizations, and analyses without writing a single line of code.