I'm pleased to announce that I've built a new harness for you - aipython. It's like ipython except you just write in English when just the right Pandas invocation escapes you
I'm a long time heavy user of ipython and I never stopped using it and neither Claude nor Claude Code solves for the same needs
- Instant answers
- And not tools or artifacts or systems, just answers. Based on data.
Having data in context is a massive advantage and having minimal goals is a great unlock - it allows me to use faster, smaller, cheaper models with great results. I have this super situational explorations many many times per day which look like
- Acquire some context (data) often in browser
- The context switch to analysis mode and refind the data
- Explore and find out
I find ipython is awesome for the explore and find out - but the middle step is a time sink. With this kind of tool I can point vaguely in the direction of the data and get the desired result. Roundtrips times are typically 1s not "thinking, discombobulating, planning, working, reporting" for minutes.
This is also my debut as a Pypi distributor, which was about time (or maybe a couple of years late depending on your views on AI)