Profiling.sampling – Statistical Profiler

Python just rolled out Tachyon, a new built-in tool in version 3.15 that watches a program from the outside and estimates where it’s spending time. In plain English: it helps developers spot what’s slowing things down without changing the code or restarting the app, which is why several commenters reacted with a very relieved “finally.” One of the loudest vibes in the room was basically, why was this not already included? A few people compared it to existing third-party tools and celebrated that Python is now shipping something like this “out of the box.”

But of course, the comment section could not simply be normal. One camp loved the convenience, while another immediately pointed out a catch: unlike Python’s older tracing tool, this one lives outside your app, so you can’t weave it into the program itself as easily. That sparked a mini philosophy battle over whether a profiler should be a built-in feature you can ship or an external detective you attach later.

Then came the classic internet detour: if Python is slow, why use Python at all? That hot take predictably invited eye-roll energy, because the whole point of Tachyon is helping people make real-world Python apps faster, not hold a language identity crisis in the comments. Meanwhile, others got practical and asked whether it can handle modern async workloads, aka the messy reality of one request juggling many tasks. So yes, the tool announcement was cool — but the real show was the community instantly splitting into the “finally!”, “isn’t this just py-spy?”, and “why are you using Python then?” factions.