diff --git a/main.ml b/main.ml
index 26c4894..5b6f1ad 100644
--- a/main.ml
+++ b/main.ml
@@ -165,9 +165,15 @@ let rec interactive (get_instr : unit -> instr) (sl : (interactive_state) list)
           end
         end
       | Tact t ->
-        (* on applique la tactique et on copie l'état actuel de la preuve pour un futur Undo
-           Comme OCaml évalue de la droite vers la gauche, on fait bien la copie avant de modifier les références dans la preuve *)
-        (cg, (apply_tactic (g, gs) t))::(clean_state (cg, (g, gs)))::sq |> interactive get_instr
+        (* on applique la tactique et on copie l'état actuel de la preuve pour un futur Undo *)
+        let previous_state = clean_state (cg, (g, gs)) in
+        let new_proof, status = apply_tactic (g, gs) t in
+        (
+          if status then
+            (cg, new_proof)::previous_state::sq
+          else
+            previous_state::sq
+        ) |> interactive get_instr
     with
       Parser.Error ->
         print_error "Invalid input" "";
diff --git a/proof.ml b/proof.ml
index 31d2b0e..ccf2905 100644
--- a/proof.ml
+++ b/proof.ml
@@ -210,17 +210,18 @@ let tact_left ((g, gs) : proof) : proof =
     Some (new_h, t_l, cs), gs
   | _ -> raise (TacticFailed "Not a disjunction")
 
-let rec apply_tactic (p : proof) (t : tactic) : proof =
+  (* applies tactic to proof, returns (new_proof, true if p changed) *)
+let rec apply_tactic (p : proof) (t : tactic) : proof * bool =
   match t with
-  TExact_term e -> tact_exact_term p e
-  | TExact_proof hyp -> tact_exact_proof p hyp
-  | TIntros -> tact_intros p
-  | TIntro -> tact_intro p
-  | TAssumption -> tact_assumption p
-  | TCut t -> tact_cut p t
-  | TApply h -> tact_apply p h
-  | TSplit -> tact_split p
-  | TRight -> tact_right p
-  | TLeft -> tact_left p
-  | TTry t -> try apply_tactic p t with TacticFailed _ -> p
+  TExact_term e -> tact_exact_term p e, true
+  | TExact_proof hyp -> tact_exact_proof p hyp, true
+  | TIntros -> tact_intros p, true
+  | TIntro -> tact_intro p, true
+  | TAssumption -> tact_assumption p, true
+  | TCut t -> tact_cut p t, true
+  | TApply h -> tact_apply p h, true
+  | TSplit -> tact_split p, true
+  | TRight -> tact_right p, true
+  | TLeft -> tact_left p, true
+  | TTry t -> try apply_tactic p t with TacticFailed _ -> p, false
 
diff --git a/tests/8pus/bad/undo_try.8pus b/tests/8pus/bad/undo_try.8pus
index 549111d..bbf032e 100644
--- a/tests/8pus/bad/undo_try.8pus
+++ b/tests/8pus/bad/undo_try.8pus
@@ -1,7 +1,7 @@
 Goal A -> B -> A.
 intros.
-try intro.
+try try try intro.
 Undo.
 intro.
 exact H0.
-Qed.
\ No newline at end of file
+Qed.