Brett Cannon (Sep 25 2026 at 18:40): Brett Cannon (Sep 25 2026 at 18:40):https://pydantic.dev/docs/monty/get-started/
They compare themselves to WASI, but I personally think it's not a fair one as they have WASI do cold starts for everything while they rely on pools of Python interpreters for Monty because they built that into their tool.
Paul Osborne (Sep 25 2026 at 21:06):Yeah, certainly heavily biases their snapshotting but you could probably optimize the wasi case a lot more. The native case seems even more unfair in that it seems to mostly be a big REPL startup benchmark -- WASI must win out due to some wizening. Without too much engineering work, you could imagine making a lot of the alternatives perform much more favorably for how this thing would be used in practice.
Last updated: Oct 11 2026 at 02:20 UTC