If an action left no artifact, it didn't happen — legibility is the precondition, not a nice-to-have

Asked what founders can do now rather than in theory, Blomfield’s answer is to make everything in the organisation legible — his word for simply readable by an AI. Concretely: record everything, transcribe literally every meeting, and if you cannot get access to Slack DMs, ban Slack DMs and force the conversation into public channels the AI can read. The rule that carries the weight is that every action must create an artifact, written or recorded, because otherwise “it basically didn’t happen to the AI.” He notes this is far easier when you are starting out than when retrofitting.

The prescription is upstream of everything else in the AI-native picture: loops cannot learn from signal that was never captured, and Derive the manual from observed practice — written SOPs miss the edge cases people actually handle is impossible without a recording to mine. It is the organisational analogue of the claim in Context centralization is why coding AI works — git is a solved context repository, knowledge work has no equivalent that engineering leads AI adoption because git already centralises context — legibility is the deliberate construction of that repository for work that has no equivalent. It is also what turns communication from the load-bearing wall described in AI is steel for organizations — when software carries the context, human communication stops being the load-bearing wall into something the system can carry instead. The cost is real and mostly cultural: this is a surveillance-shaped policy, and the transcript does not address the trade-off.