At Exafluence, it started with the desire to get LLMs to do what they are not good at: computation. LLM foundation models are great tools for creative and generative tasks. They do their best when asked to write a poem or a story. However, they fall short of expectations in computation-intensive use cases. The power of LLMs in interpreting natural language from human users was difficult to ignore. We saw great potential in strengthening LLMs with computational capabilities. We zeroed in on the Wolfram Engine for that purpose.