Worked on the jupyterlab/jupyter-ai repository to address a reliability issue with AI-generated code insertion in IPython terminals. Focused on improving the integration between JupyterLab and IPython by replacing the use of payload_manager.write_payload with self.shell.set_next_input, ensuring that code generated by AI is correctly handled in various IPython environments. This change preserved the expected behavior within JupyterLab while enhancing stability for users working in notebook and terminal contexts. The work involved Python development and a deep understanding of both IPython and Jupyter internals, resulting in a targeted bug fix that improved the robustness of AI-assisted coding workflows.