We Cut 94% of AI Coding Tokens With a Local Code Index - Rajkumar Sakthivel, Tesco
A local code-indexing search layer cut AI coding input tokens by 94%, saving ~61% of total cost.
“It's like ordering a pizza and paying for extra nine pizzas you don't eat every time.”
A Tesco engineer described building a local code index that retrieves only relevant code snippets instead of sending whole files as context, cutting input tokens by 94% and total AI coding costs by ~61%. The talk highlights that ~90% of AI coding spend is input context, not output, making input reduction the highest-leverage optimization. It matters as a practical cost-control pattern for teams scaling AI-assisted development.