Shopify's Tangle lets machine-learning experiments reuse intermediate results instead of repeating identical computation. Researchers connect processing steps into a visual pipeline, and Tangle tracks the files each step produces. It compares file contents using hashes, compact identifiers derived from those contents. If an upstream change produces identical output, later steps can reuse their cached results. Teams can also share a matching computation that is still running. The design makes repeated experiments cheaper to execute when they share unchanged work.
