Software engineering calories

Using a calorie-tracking analogy, this piece argues that most engineering teams have no idea how much code (input) they're actually shipping relative to feature adoption (output). It explores why simple metrics like lines of code or PR counts are poor proxies, advocates for ML-based tools like Weave to estimate the real 'caloric value' of engineering work, and cautions that more code isn't always better — some products would benefit from deleting features rather than adding them. The author also warns that AI-generated code may look productive in the short term but degrade product health over time.