Airtable views enable users to share filtered data with customers, partners, and others and easily consume relevant information. To deliver views efficiently, Airtable developed the “Recalculator” which analyzes write operations to assess what formulas, filters, and sorts should be re-evaluated when a user interacts with a view. In this post, Keyhan Vakil discusses how Airtable developed a fuzzer to find Recalculator bugs by performing actions like adding and deleting columns or rows. He discusses how the fuzzer generates valid CRUD requests while updating the state, explores requests that generate new code coverage, and detects errors that don’t result in crashes.