Fellowship of the Stack: Scientific Discovery with Python
summary
A practical tour of scientific Python that names tools, tradeoffs and numbers: a Python packaging survey has 16,000 respondents and closes end of December; environment tools like Conda, Mamba, pip, poetry and Twine are compared; performance wins highlight Polars 1.0 (order of magnitude faster) and GPU work like Rapids up to 13× speedups; packaging headaches include GDAL, Pillow and PyTorch ABI mismatches; recommendations cover reproducible environments, CI/CD, linters/Bandit, PEPs (440, 621, 518) and PSF membership options.
AI-written summary · about
chapters (26)
- 0:00Introduction and speaker intro
- 1:51Talk overview and narrative arc
- 5:58Why Python ecosystem is treacherous
- 8:10Evolving nature of software
- 10:53Current ecosystem improvements
- 11:37Forming the scientific Python fellowship
- 14:30Interactive dependency discussion
- 16:38Scientific companion and ecosystem tools
- 18:08First generation scientific Python tools
- 20:24New magic in scientific Python
- 21:30Python packaging survey and PEP Open SI
- 22:16Python as a glue language
- 23:40C Python interpreter deep dive
- 28:23Packaging complexities and challenges
- 31:44Domain expertise and package selection
- 32:48Python distributions and ecosystem trends
- 35:03Scaling your quest with code
- 37:10Scaling your quest with the PSF
- 38:13PSF funding and community support
- 40:21Analogy limitations and closing remarks
- 41:46Q&A: PSF membership age
- 42:28Q&A: Packaging tools and PEPs
- 43:58Q&A: Python Asia and PSF relation
- 45:48Q&A: Python concurrency improvements
- 47:16Q&A: Python's future and threats
- 48:45Q&A: WebAssembly and Python
