Shed Skin restricted-Python to C++ transpiler v1.0 released The Shed Skin project released version 1.0 of its restricted-Python to C++ transpiler after more than 20 years of development. The maintainer credits a much improved type inference engine, Unicode support, faster generated code, Python 3.15 compatibility, and AI assistance (Claude) that helped identify and often fix roughly 500 bugs. The project reports generated code now runs on average 5.0 times faster than PyPy in its own benchmark plot. I have just released version 1.0 https://github.com/shedskin/shedskin/releases/tag/v1.0 of Shed Skin https://github.com/shedskin/shedskin , a restricted-Python to C++ transpiler. It has taken a long time, in fact more than 20 years, to get to this point. I want to start by thanking all the people that have contributed over time. Leonardo Maffi helped out by testing in the first few years. IIRC, Harri Pasanen contributed extension module support. Johan Kristensen was instrumental in the move to Python 3. Jeremy Roquet and others helped to keep the project alive in my absense of several years. And in recent years, Shakeeb Alireza has been the main contributor, adding CMake and hence good Windows support, cleaning up the code and tests, and also contributing many good ideas. So why release 1.0 now? Well, let's say that many things have finally come together. From a much improved type inference engine, to Unicode support, to much improved performance of generated code, to basically catching up with Python 3.15, a lot has happened over the past few months. Another thing that has happened over the past few years? is that C++ compilers have started to optimize away object allocations where possible. This has always been an achilles heel of Shed Skin, making it slower than for example PyPy for certain allocation-heavy programs. Regular allocations also seem to have become faster over time under Linux at least . For my standard of course highly biased performance plot, Shed Skin is now on average 5.0 times faster than PyPy, and in no case slower. Something that I have to acknowledge also is the help of AI, more specifically Claude, to develop Shed Skin at a much more rapid speed than possible before. It has also been able to identify and in many cases fix about 500 bugs. We are all wondering about the future impact of AI of course, but let's hope that programming will remain a human hobby for a long time the way chess is still enjoyable for humans, even though we totally suck at it.. :- . But AI doesn't only help with development of Shed Skin, it can also help users to make their code Shed Skin compatible, and/or even submit issues/contributions. The following new example an 'Another World' engine , was changed by AI in a few places, so that it worked with Shed Skin while of course remaining valid CPython code, with no added type annotations : Thanks again to everyone who has contributed to the project, which has turned out to be my "life's work". Now that 1.0 has been reached, and with the advent of AI in programming, I don't know if I will continue to work on it intensively anymore though, without more/any community feedback or financial support, and spend more of my free time on other interesting projects. I will certainly try to keep it working, and do a yearly update for every new CPython release. I probably won't be able to help myself, but we will see :-