Introduction

Recent Python releases (3.12+) on Windows introduced a much cleaner, first-class workflow centered around the Python Launcher (py) and improved packaging tools. This post documents a modern, low-friction Python workflow that works cleanly with Python 3.14+ and scales forward.

This is not about experimental tooling — it reflects officially supported behavior from Python.org and Microsoft.

Wiping Old Python Installations (Recommended)

If you have experimented with Python over the years, your system may contain multiple abandoned installs, broken PATH entries, or orphaned pip --user packages.

Before adopting the modern workflow, it is often best to start clean.

I provide a PowerShell script that completely removes legacy Python installations, including:

  • Old python.org installs
  • Microsoft Store Python
  • Orphaned pip --user packages
  • Leftover registry entries
  • Stale PATH references

nuke_python.ps1

Run it from an elevated PowerShell session, reboot, then install Python fresh from python.org.

The Python Install Manager on Windows

On Windows, Python is no longer managed by juggling python.exe paths.

Instead, the Python Launcher (py) acts as the install manager and version selector.

Key characteristics:

  • Installed automatically by the python.org installer
  • Knows about all installed Python versions
  • Selects the correct interpreter without PATH hacks
  • Future-proof across Python upgrades
py list
py -3.14

Installing and Updating pip, setuptools, and wheel

py -3.14 -m pip install -U pip setuptools wheel

This command updates the entire Python packaging toolchain for the selected interpreter.

Scripts, Entry Points, and Where Commands Come From

When a package installs a command-line tool, it is placed in the Scripts directory of the active environment.

Virtual environments keep these tools isolated per project.

Example: Installing and Using pdf2docx

py -3.14 -m venv .venv
.venv\Scripts\Activate.ps1
python -m pip install pdf2docx
pdf2docx input.pdf output.docx
  1. Wipe legacy installs (optional but recommended)
  2. Install Python from python.org
  3. Use py to select versions
  4. Use venvs per project
  5. Install dependencies only inside the venv

Closing Thoughts

Python on Windows has finally converged on a boring, reliable workflow — and that is a good thing.