diff --git a/use-based-refs/interpreter/README.md b/use-based-refs/interpreter/README.md
index fd4d6c557ccbf22b6b82b0fb8b97f204fd4bb84b..f5312b9f53e0e8ce116365a1402a86c863734027 100644
--- a/use-based-refs/interpreter/README.md
+++ b/use-based-refs/interpreter/README.md
@@ -14,6 +14,7 @@ Right now, the interpreter only executes a program hardcoded in `main`. Change t
 
 - [ ] Input files/nice CLI
 - [ ] Pretty error messages (using miette/ariadne)
+    - [ ] For this, the AST should keep track of Spans
 - [ ] Unify test cases for typechecking and evaluating, where possible
 - [ ] Language features
     - [x] `TypedIdent`
@@ -23,12 +24,13 @@ Right now, the interpreter only executes a program hardcoded in `main`. Change t
     - [ ] Allow for `&*` reborrowing
     - [ ] Function definitions
     - [ ] Function calls
-    - [ ] Allocations
-    - [ ] Dereference -- read
+    - [x] Allocations
+    - [x] Dereference -- read
     - [ ] Dereference -- write
 
 ## Design decisions to make
 
 - [ ] Should type annotations be possible for any expression? For any identifier? Currently, they are only allowed on the LHS of let expressions.
 - [ ] Right now, allocation of uninitialized memory is not possible; should that be changed?
-- [ ] Should blocks be able to be (de-)referenced à la `&{let x = 42; x}`? Probably yes, but there seem to be problems with the implementation. Similarly: What about let expressions etc.?
\ No newline at end of file
+- [ ] Should blocks be able to be (de-)referenced à la `&{let x = 42; x}`? Probably yes, but there seem to be problems with the implementation. Similarly: What about let expressions etc.?
+- [ ] Switch to typed AST? Right now, the evaluator can run independent of the typechecker; this may lead to panics/weird behavior if the program is parseable but doesn't typecheck. Are there any advantages to requiring typecking before evaluation?
\ No newline at end of file