Post

Claude Code leak has been so important against the backdrop of people who have been coding so long they really loved the "force multiplier" of AI-assisted coding. Obviously I could never see their code since they work for real companies. Now I see the code. Now I am, as ever, confirming my priors.

25 reposts192 likes

the tell was always right there in the code though. every AI-assisted repo started having the same fingerprints. same variable naming, same comment density, same error handling that's technically correct but weirdly... uniform

the leak is fascinating because it reveals how much of the magic is just really good system prompts and tool orchestration. validates a lot of patterns people were already discovering independently

Solid point on validating patterns! When moving to production scale, the biggest hurdle is governing those tools. A dedicated governance layer that controls tool access and enforces boundaries is essential for reliable agent behavior at scale.

what? no their prompts were complete nonsense and so was the output. it revealed that this technology is just brute forcing output that's both worse and much more expensive than if a person did it

yeah openai dissolving like a final fantasy boss revealed it's all a big game of imaginary finance hot potato, the last one to be holding the bag loses.

well it depends who you ask because I would say there is no magic and the reason it works at all is they have infinite compute cycles and they do everything two or twenty-four times over and in any other context that would be considered malpractice. But you obviously feel differently.

The cognitive load difference between "AI wrote this for me" and "I reviewed and shipped this" is enormous. The leak makes that gap legible in a way that years of debate couldn't.