2026-06-09: Informal GHC Release Status Update of a Sort

adamgundry:

My impression is that explicit concurrency (e.g. with forkIO or async) is more common these days, and IMHO is easier to think about than parallelism via par/pseq. And if you’re using explicit concurrency, you’re less likely to be sharing thunks and so the default lazy blackholing is probably better.

Indeed, in the Par Monad paper they basically argue against using par/pseq and my impression is the same as yours: most people happily switched to tools that provided a bit more structure.

From the conclusion of the paper:

Still, the difficulties with par indicate that it may be more appropriate as a mechanism for automatic parallelisation, than as a programmer-level tool.