redsquirrel12

I built it to explore a few things:

1. how to help analysts stay on top of the analysis when agents can output it faster than they think

2. how much of the sensitive analysis/workflow augmentation can be done using local models on standard machines (targeting 2026 laptops) with only the summaries leaving the computer to be processed by a frontier

3. what interactive computing UI such as Jupyter should offer under assumption that coding is solved by AI, but human understanding is the bottleneck

Repository: https://github.com/openteams-ai/whybook

I spent a while making JupyterLab's visual regression tests less painful to work with - for both humans and coding agents.

This covers what we changed, what did not work, and what we learned about integrating visual regression testing into the modern development workflow.

This is extremely annoying for maintainers who are trying to review PRs and just cannot find stuff. It's been going for over a month and GitHub is ignoring it. Not to mention 500 pages when you try to merge security fixes or report spam.

AFAIK they have not even tried contributing back to Jupyter.

At the time when job openings are declining across the board, it's easy to lie with statistics.

It's like publications saying "research discipline X is gaining traction in recent years" based on search matches in PubMed, but not normalizing for the total number of articles submitted increasing.