Several of this week’s top stories touch on the pitfalls of open source development, especially when things like power, money, and ego are involved. Also, a look at Python’s nifty new sampling profiler, and a play-by-play account of using Qwen3.5 on a Python project. Credit: ozalpvahid / Shutterstock Announced at JavaOne, Project Detroit proposes to break down the walls between Java, Python, and JavaScript. Also in this report: Better ways to instrument your code with Python’s new built-in sampling profiler, another run at using AI locally to rework a Python project, and the question on everyone’s mind right now (surely): What does OpenAI really want with Astral? Top picks for Python readers on InfoWorld OpenAI buys Python tools builder AstralAstral, the maker of uv, ty, and pyx, has a new home under the OpenAI umbrella. Is OpenAI demonstrating its commitment to maintaining tooling in the AI space, or is the purchase more of a power move? I ran Qwen3.5 locally instead of Claude Code. Here’s what happenedWant to run an LLM on your own hardware for that at-home Claude Code or Copilot experience? You can, but it’ll be a bumpy ride. My takeaway? Maybe don’t let the AI run around unsupervised after dark. Hands-on with the new sampling profiler in Python 3.15Among Python 3.15’s best new features is a sampling profiler. See how it works in this guide to using the profiler to instrument your code and find bottlenecks with minimal performance impact. Project Detroit, bridging Java, Python, JavaScript, moves forwardThe once-dead, now-revived Detroit project aims to allow Java’s Foreign Function and Memory API to talk seamlessly to other language runtimes. The vision? More powerful mixing and matching of languages across domains. More good reads and Python updates elsewhere The slow collapse of MkDocsThe strange, ongoing saga of how a developer meltdown took out one of the most popular documentation tools for Python—with no clear successor in sight. Comparing the typing spec conformance of Python type-checking toolsHow well do tools like Pyright, Pyrefly, Mypy, Ty, and others conform to Python’s own type annotation specs? The answers range, surprisingly, from “very closely” to “just barely.” The optimization ladder: All the ways to make Python fasterFrom replacing the runtime to integrating modules written in C or Rust, here’s an end-to-end rundown of ways to speed up Python for tasks that urgently need performance. License laundering and the death of ‘clean room’When someone rewrote a long-unmaintained Python library with an LLM, the original developer broke a decade-plus silence to object. What are the implications for open source? Programming LanguagesPythonSoftware DevelopmentJavaArtificial IntelligenceOpen SourceDevelopment ToolsCareers