diff --git a/R/PipeOp.R b/R/PipeOp.R
index 3da95e78..e98cbfa6 100644
--- a/R/PipeOp.R
+++ b/R/PipeOp.R
@@ -61,11 +61,11 @@ PipeOp = R6::R6Class("PipeOp",
       private$.id = id
       private$.param_set = param_set
       #FIXME: we really need a function in paradox now to get defaults
-      private$.param_vals = param_set$data$default
-      names(private$.param_vals) = param_set$ids
-      private$.param_vals = insert_named(private$.param_vals, param_vals)
-      if (!param_set$test(private$.param_vals)) {
-        stop("Parameters out of bounds")
+      private$.param_vals = param_set$defaults
+      if(!is.null(private$.param_vals)) {
+        if (!param_set$test(private$.param_vals)) {
+          stop("Parameters out of bounds")
+        }
       }
     },
 
diff --git a/R/PipeOpDT.R b/R/PipeOpDT.R
index d292add1..9b14a2b6 100644
--- a/R/PipeOpDT.R
+++ b/R/PipeOpDT.R
@@ -50,10 +50,10 @@ PipeOpDT = R6Class("PipeOpDT",
       list(TaskClassif$new(id = task$id, backend = db, target = tn))
     },
 
-    predict = function() {
-      assert_list(self$inputs, len = 1L, type = "Task")
+    predict = function(inputs) {
+      assert_list(inputs, len = 1L, type = "Task")
       assert_function(self$predict_dt, args = "newdt")
-      task = self$inputs[[1L]]
+      task = inputs[[1L]]
       fn = task$feature_names
       d = task$data()
 
@@ -66,7 +66,9 @@ PipeOpDT = R6Class("PipeOpDT",
       d[, (colnames(dt)) := dt]
       d[, "..row_id" := seq_len(nrow(d))]
 
-      list(task$overwrite(d))
+      db = DataBackendDataTable$new(d, primary_key = task$backend$primary_key)
+      tn = task$target_names
+      list(TaskClassif$new(id = task$id, backend = db, target = tn))
     }
   )
 )
diff --git a/R/PipeOpLearner.R b/R/PipeOpLearner.R
index ea7db536..543687a2 100644
--- a/R/PipeOpLearner.R
+++ b/R/PipeOpLearner.R
@@ -40,9 +40,10 @@ PipeOpLearner = R6Class("PipeOpLearner", inherit = PipeOp,
       # private$.result
     },
 
-    predict2 = function() {
+    predict = function(inputs) {
       assert_list(inputs, len = 1L, type = "Task")
-      predict(self$state)
+      task = inputs[[1L]]
+      list(self$state$learner$predict(task))
     }
   ),
 
diff --git a/tests/testthat/test_usecases.R b/tests/testthat/test_usecases.R
index 637894bf..5714e7c4 100644
--- a/tests/testthat/test_usecases.R
+++ b/tests/testthat/test_usecases.R
@@ -4,10 +4,17 @@ test_that("scale + pca", {
   task = mlr_tasks$get("iris")
   g = PipeOpScale$new() %>>% PipeOpPCA$new()
   res1 = g$train(task)
-  assert_list(res1)
   res2 = g$predict(task)
 
 })
 
+test_that("scale + pca + PipeOpLearner", {
+  task = mlr_tasks$get("iris")
 
+  g = PipeOpScale$new() %>>%
+    PipeOpPCA$new() %>>%
+    PipeOpLearner$new(mlr_learners$get("classif.rpart"))
+  res1 = g$train(task)
+  res2 = g$predict(task)
+})