diff --git a/AGENTS.md b/AGENTS.md
index 7657bb2..649a676 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -3,7 +3,7 @@
 R package for **partial dependence plots (PDPs)** and **individual conditional
 expectation (ICE) curves** from fitted ML models. Exports: `partial()` (the
 workhorse), `plot()` methods (tinyplot/base graphics, default engine),
-`plotPartial()` (lattice), `exemplar()`, and the deprecated `topPredictors()`.
+`plotPartial()` (lattice), and `exemplar()`.
 
 ## Branches & releases
 
@@ -12,8 +12,10 @@ workhorse), `plot()` methods (tinyplot/base graphics, default engine),
 - **`main`**: stable releases only, tagged `vX.Y.Z`. r-universe
   (pinned to main in bgreenwell/bgreenwell.r-universe.dev) and the pkgdown
   site both build from main — never push experimental work there.
-- Release: merge devel → main, drop the `.9000` suffix and dev NEWS heading,
-  tag, push; then bump devel to the next `.9000`.
+- Release: merge devel → main (`--no-ff`), drop the `.9000` suffix and dev
+  NEWS heading, tag, push, `gh release create`; then **merge main back into
+  devel** (else the release merge commit leaves devel "behind" main) and bump
+  devel to the next `.9000`.
 - Shared fixes that main needs immediately (CI, README): commit to **main
   first, then merge main → devel**. Never cherry-pick devel → main — it
   duplicates commits and makes main appear "ahead" of devel.
diff --git a/DESCRIPTION b/DESCRIPTION
index bd2f589..ed3d44d 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -1,7 +1,7 @@
 Package: pdp
 Type: Package
 Title: Partial Dependence Plots
-Version: 0.9.1
+Version: 0.9.1.9000
 Authors@R: person(c("Brandon", "M."), family = "Greenwell",
     email = "greenwell.brandon@gmail.com", role = c("aut", "cre"),
     comment = c(ORCID = "0000-0002-8120-0084"))
diff --git a/NAMESPACE b/NAMESPACE
index dc21770..4b21a92 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -11,12 +11,9 @@ S3method(plot,partial)
 S3method(plotPartial,cice)
 S3method(plotPartial,ice)
 S3method(plotPartial,partial)
-S3method(topPredictors,default)
-S3method(topPredictors,train)
 export(exemplar)
 export(partial)
 export(plotPartial)
-export(topPredictors)
 export(trellis.last.object)
 importFrom(lattice,dotplot)
 importFrom(lattice,equal.count)
diff --git a/NEWS.md b/NEWS.md
index 2675a74..0c51236 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,3 +1,12 @@
+# pdp (development version)
+
+## Breaking changes
+
+* Removed `topPredictors()`, which had been deprecated since pdp 0.8.0; use
+  the [vip](https://bgreenwell.github.io/vip/) package to identify important
+  predictors instead.
+
+
 # pdp 0.9.1
 
 ## Deprecations
diff --git a/R/topPredictors.R b/R/topPredictors.R
deleted file mode 100644
index a78e03e..0000000
--- a/R/topPredictors.R
+++ /dev/null
@@ -1,88 +0,0 @@
-#' Extract Most "Important" Predictors (Experimental)
-#'
-#' Extract the most "important" predictors for regression and classification
-#' models.
-#'
-#' @param object A fitted model object of appropriate class (e.g., `"gbm"`,
-#' `"lm"`, `"randomForest"`, etc.).
-#'
-#' @param n Integer specifying the number of predictors to return. Default is
-#' `1` meaning return the single most important predictor.
-#'
-#' @param ... Additional optional arguments to be passed onto
-#' [caret::varImp()].
-#'
-#' @details
-#' This function uses the generic function [caret::varImp()] to
-#' calculate variable importance scores for each predictor. After that, they are
-#' sorted at the names of the `n` highest scoring predictors are returned.
-#'
-#' @rdname topPredictors
-#' @export
-#' @examples
-#' \dontrun{
-#' #
-#' # Regression example (requires randomForest package to run)
-#' #
-#'
-#' # Load required packages
-#' library(randomForest)
-#'
-#' # Fit a random forest to the mtcars dataset
-#' data(mtcars, package = "datasets")
-#' set.seed(101)
-#' mtcars.rf <- randomForest(mpg ~ ., data = mtcars, mtry = 5, importance = TRUE)
-#'
-#' # Top four predictors
-#' top4 <- topPredictors(mtcars.rf, n = 4)
-#'
-#' # Construct partial dependence functions for top four predictors
-#' pd <- NULL
-#' for (i in top4) {
-#'   tmp <- partial(mtcars.rf, pred.var = i)
-#'   names(tmp) <- c("x", "y")
-#'   pd <- rbind(pd,  cbind(tmp, predictor = i))
-#' }
-#'
-#' # Display partial dependence functions
-#' tinyplot::tinyplot(y ~ x, facet = ~ predictor, data = pd, type = "l",
-#'                    facet.args = list(free = TRUE), ylab = "mpg")
-#'
-#' }
-topPredictors <- function(object, n = 1L, ...) {
-  .Deprecated(
-    msg = paste("pdp::topPredictors() is now deprecated and will be removed in",
-                "the next version. For a more general replacement, check out",
-                "the vip package: https://github.com/koalaverse/vip.")
-  )
-  if (!requireNamespace("caret", quietly = TRUE)) {
-    stop("Package \"caret\" is needed for this function to work. Please ",
-         "install it.", call. = FALSE)
-  }
-  UseMethod("topPredictors")
-}
-
-
-#' @rdname topPredictors
-#' @export
-topPredictors.default <- function(object, n = 1L, ...) {
-  imp <- caret::varImp(object, ...)
-  if (n > nrow(imp)) {
-    n <- nrow(imp)
-  }
-  imp <- imp[order(imp[, "Overall"], decreasing = TRUE), , drop = FALSE]
-  rownames(imp)[seq_len(n)]
-}
-
-
-#' @rdname topPredictors
-#' @export
-topPredictors.train <- function(object, n = 1L, ...) {
-  # FIXME: What about filterVarImp?
-  imp <- caret::varImp(object, ...)$importance
-  if (n > nrow(imp)) {
-    n <- nrow(imp)
-  }
-  imp <- imp[order(imp$Overall, decreasing = TRUE), , drop = FALSE]
-  rownames(imp)[seq_len(n)]
-}
diff --git a/_pkgdown.yml b/_pkgdown.yml
index f8c88cf..07333cc 100644
--- a/_pkgdown.yml
+++ b/_pkgdown.yml
@@ -38,6 +38,3 @@ reference:
   contents:
   - boston
   - pima
-- title: Deprecated
-  contents:
-  - topPredictors
diff --git a/man/topPredictors.Rd b/man/topPredictors.Rd
deleted file mode 100644
index 02086c5..0000000
--- a/man/topPredictors.Rd
+++ /dev/null
@@ -1,64 +0,0 @@
-% Generated by roxygen2: do not edit by hand
-% Please edit documentation in R/topPredictors.R
-\name{topPredictors}
-\alias{topPredictors}
-\alias{topPredictors.default}
-\alias{topPredictors.train}
-\title{Extract Most "Important" Predictors (Experimental)}
-\usage{
-topPredictors(object, n = 1L, ...)
-
-\method{topPredictors}{default}(object, n = 1L, ...)
-
-\method{topPredictors}{train}(object, n = 1L, ...)
-}
-\arguments{
-\item{object}{A fitted model object of appropriate class (e.g., \code{"gbm"},
-\code{"lm"}, \code{"randomForest"}, etc.).}
-
-\item{n}{Integer specifying the number of predictors to return. Default is
-\code{1} meaning return the single most important predictor.}
-
-\item{...}{Additional optional arguments to be passed onto
-\code{\link[caret:varImp]{caret::varImp()}}.}
-}
-\description{
-Extract the most "important" predictors for regression and classification
-models.
-}
-\details{
-This function uses the generic function \code{\link[caret:varImp]{caret::varImp()}} to
-calculate variable importance scores for each predictor. After that, they are
-sorted at the names of the \code{n} highest scoring predictors are returned.
-}
-\examples{
-\dontrun{
-#
-# Regression example (requires randomForest package to run)
-#
-
-# Load required packages
-library(randomForest)
-
-# Fit a random forest to the mtcars dataset
-data(mtcars, package = "datasets")
-set.seed(101)
-mtcars.rf <- randomForest(mpg ~ ., data = mtcars, mtry = 5, importance = TRUE)
-
-# Top four predictors
-top4 <- topPredictors(mtcars.rf, n = 4)
-
-# Construct partial dependence functions for top four predictors
-pd <- NULL
-for (i in top4) {
-  tmp <- partial(mtcars.rf, pred.var = i)
-  names(tmp) <- c("x", "y")
-  pd <- rbind(pd,  cbind(tmp, predictor = i))
-}
-
-# Display partial dependence functions
-tinyplot::tinyplot(y ~ x, facet = ~ predictor, data = pd, type = "l",
-                   facet.args = list(free = TRUE), ylab = "mpg")
-
-}
-}