date: mid-august (free-threading push)

wanted:

• reducing the number of objects per release by 25%: no more"whl" version
• wppm usable in jupyterlite and xeus: from wppm import piptree;pip = piptree.PipData();print(pip.down('.'))
• modernize infrastructure to be compatible with Python-3.16 and Trusted Publishing
• numba-0.66.0, but so no numpy-2.5.0 yet
• using "paid" CLAUDE Code to simplify maintenance and improve quality
• ai local alpha but getting clother, via https://github.com/microsoft/onnxruntime-genai, but Phi-3-mini can't be included ... but Google chrome does download its own model... so a question of when-not-if
• no Python-3.15.0rc1 [60% of] slim with available binary packages... not the practical level of 80%

remark:

• python-3.15 free-threading should get massively rapid adoption, because of speed-up, abi3.abi3t, 50% of users, 80% of dev being AI
• ... example: scaling free-threading 32x (vs 16x) on numpy+cpython-3.15 three months before release : 30x faster vs 5x via multi-processing
• for further speed, bits of rust will be incorporated in cpython around 3.17: "Python interface, Rust engine"
• jupyterlite / pyodide is getting more flexible, because of PEP-783 allowing pypi wheels
• scikit-learn starting to support gpu acceleration
• new Windows "Python Install manager" is progress
• python at 50% of pypl.github.io (tutorial search) in august 2026: guess is 30% of humans, 70% of AIs ... so why not 60% at christmas