Henrik Bengtsson presented a talk at the useR! 2026 conference in Warsaw, Poland, introducing the new futurize package, part of the futureverse ecosystem. The package lets R users parallelize existing map-reduce style calls (lapply, purrr::map, foreach, plyr::llply) simply by piping them to futurize(), and it also supports domain-specific functions like stats::kernapply, boot, caret::train, glmnet::cv.glmnet, and lme4::allFit. Slides (16 slides, 18 minutes) are linked, with video to appear. The post also notes next year's useR! conference will be held in Santiago, Chile in July 2027.
Table of contents
RelatedQuestions this post answers
How do I parallelize an lapply call in R without rewriting it as a future_lapply call?
Pipe the existing call to the futurize() function from the new futurize package. For example, ys <- lapply(xs, fit_model) |> futurize() parallelizes the call automatically. The same pattern works for purrr::map, foreach %do%, and plyr::llply calls, requiring no rewrite of the underlying logic. Developers weighing parallelization approaches in R can follow futureverse updates like this on daily.dev.
Which R modeling functions can be parallelized with the futurize package?
The futurize package supports domain-specific calls beyond basic map-reduce functions, including stats::kernapply, boot::boot, caret::train, glmnet::cv.glmnet, and lme4::allFit. Each can be parallelized simply by piping the call to futurize(), for example model <- caret::train(Species ~ ., data = iris, method = "rf", trControl = ctrl) |> futurize(). Anyone speeding up R model training workflows can track futurize package updates on daily.dev.