Very performant, very easy to distribute cross-compiled static binaries, very portable to web, embedded, anywhere.
The main downside of Rust, being that it has a high learning curve and is costly to write in because of all the training, could be less and less of a problem.
Compiled managed languages are fast enough for all kinds of workloads I work on.
Also I am polyglot since the days of Timex 2068, so a little bit of FFI isn't an issue. No need to throw away the productivity tooling of the baby with the water.
Finally, as someone coding since the 8 bit home computer days, the religious discussion of static linking belongs into the same basket as using an i9 with a NVidia RTX to power a UNIX System V like command line experience, or reliving the glory days of curses, Clipper, FoxPro or Turbo Vision applications in 2025.
One can get too obsessed about performance.
Productivity with AI-generated Rust should be on par with anything since the verbosity is only measured in tok/s.
I am already aware of at least one project where the contractor was doing the required assignment in a Python prototype, and then the client team used AI tooling to convert it into Go for the actual production code.
Even this, will be more for debugging purposes, as nothing prevents to use the same prompts to eventually generate Assembly or machine code directly.
Sure, LLMs are still a bit buggy for that in 2025, nonetheless not imagining how programming will look in 10 years time, with the hindsight that we are only at this AI hype cycle for two years, will be the failure of those that can only identity themselves with being a technology XYZ developer.
Just imagine the case, when you may save on a cloud several hundred thousand dollars per year just choosing more effective programming language.