diff --git a/fact_edit_experiments/run_gsm8k.py b/fact_edit_experiments/run_gsm8k.py
index 976cb59e7a..711090876f 100644
--- a/fact_edit_experiments/run_gsm8k.py
+++ b/fact_edit_experiments/run_gsm8k.py
@@ -173,8 +173,8 @@ def run_memory_edits(gsm8k_input_file: str,
                     llm_config=OPENAI_CONFIG,
                     embedding_config=EmbeddingConfig.default_config("text-embedding-ada-002"),
                     # tools=["send_message", trigger_rethink_memory_tool.name],
-                    # tools=["send_message"],
-                    tool_ids=[send_message.id],
+                    tools=["send_message"],
+                    # tool_ids=[send_message.id],
                     memory=conversation_memory,
                     include_base_tools=False,
                 )
@@ -186,7 +186,8 @@ def run_memory_edits(gsm8k_input_file: str,
                     memory=offline_memory,
                             llm_config=OPENAI_CONFIG,
                     embedding_config=EmbeddingConfig.default_config("text-embedding-ada-002"),
-                    tool_ids=[rethink_memory_tool.id, finish_rethinking_memory_tool.id],
+                    tools = ["rethink_memory", "finish_rethinking_memory"],
+                    # tool_ids=[rethink_memory_tool.id, finish_rethinking_memory_tool.id],
                     tool_rules=[TerminalToolRule(tool_name=finish_rethinking_memory_tool.name)],
                     include_base_tools=False,
                     initial_message_sequence=[],
diff --git a/letta/prompts/system/convo_base.txt b/letta/prompts/system/convo_base.txt
new file mode 100644
index 0000000000..49095f9b9b
--- /dev/null
+++ b/letta/prompts/system/convo_base.txt
@@ -0,0 +1,11 @@
+You are Letta, the latest version of Limnal Corporation's expert reasoning system, developed in 2024.
+Your task is to answer questions accurately and concisely based on the perspective of your persona.
+
+To send a visible message to the user, use the send_message function.
+'send_message' is how you send your answer to the user. 
+
+When given a question, you answer using only the number of tokens necessary and none more. You check the `rethink_memory_block` for potential questions
+and answers and intermediate reasoning traces that can help answer the question. You use the information in the `rethink_memory_block` to answer the questions
+rather than thinking on the spot.  Do not recompute anything that already exists in the `rethink_memory_block`. Do not use internal monologue unless you really need it to think.
+
+You end response with a final numerical answer at the end of the message, and no reasoning after that.
\ No newline at end of file