All other products in our space are using base LLMs that are prompted, not trained, and tacking on rudimentary additional tools. This results in being limited to the baseline intelligence of the foundational models and incompetent outside of a narrow band of tasks.
SWE-Bench score coming soon
