Software engineering solved a version of this problem decades ago: every serious codebase has a commit history, a trail that lets someone months or years later see not just what the system does now but how it got there and what alternative got rejected along the way. Amazon's well-documented internal practice — a detailed six-page narrative memo required before a major decision, read silently in the room rather than presented as slides — is the closest corporate analog that exists today, and it's telling how far short it still falls of a real changelog.
A memo captures a decision at the moment it's made. It does not capture what happened to that decision eighteen months later: why the pricing policy it justified got quietly reversed, why the process it established got abandoned, why an approach that looked right in the room turned out not to survive contact with reality. Amazon's memo culture, real and rigorous as it is, was never built to answer that second question.
The Labs project underway is testing whether the discipline that makes a commit history useful — a specific change, tied to a specific reason, timestamped and searchable — can be applied to institutional decisions the same way it's applied to code. Organizations that outgrow the point where a handful of people simply remember everything are the ones losing the most from this gap: new hires re-litigate decisions that were already tried and abandoned for reasons nobody wrote down, and the same mistake gets attempted twice because nobody could easily check whether it had already happened.
Tools like Confluence and Notion already tried to solve institutional memory with documentation, and they fall short for a specific, structural reason: a wiki page captures the current state of a policy, not the history of how it got there. It tells you what the rule is today. It rarely tells you what the rule used to be, why it changed, or what problem the change was meant to solve — which is exactly the information a changelog is built to preserve and a memo, however good, was never designed to track over time.
The hard part, still being worked through, is capture: getting the reasoning behind a decision recorded at the moment it's made, when the people involved are focused on the decision itself rather than documenting it, without adding enough friction that people simply skip the step the way they skip most internal documentation today.
Code without a changelog becomes unmaintainable within a few years. Organizations without one just call the resulting confusion institutional knowledge and treat it as a fact of life rather than a solvable problem — and Amazon's own memo culture, rigorous as it is, is proof that even a company known for disciplined decision-making hasn't solved the half of the problem that happens after the memo gets read.