Riffing on this tweet from Dax about how ideas being expensive to implement was actually doing us a favor.
I also watched a video recently from Thorsten Ball where he stated I am the bottleneck now. Thorsten's story described a customer who had filed a well-written bug report and he didn't see it for hours. When he did, he pasted it into GPT-5.3 Codex and the agent produced a fix that he reviewed and merged. His realization was: I'm now the bottleneck. If the agent could have acted immediately, the fix would have shipped hours sooner. The kicker is, the customer didn't actually investigate anything either. They'd used an AI agent to file the bug report. The entire loop was AI-to-AI with a human in the middle slowing it down. Thorsten went on to say that the entire stack of software around project management today is built with the axiom that ideas are expensive, and that this axiom is no longer true, both humans and software need to evolve to account for this new reality.
But being the bottleneck is good. If you let agents autonomously file bugs, fix them, and deploy, you're effectively giving customers carte blanche to direct your product. That scales to slop fast (see: projects like moltbook, 4claw, and Gas Town).
Dax is saying that when ideas were expensive to implement, it forced prioritization and saying no. Cheap implementation means bad ideas ship just as fast as good ones. Good engineers, PMs, and designers will get progressively frustrated watching product quality degrade, and they'll probably choose to leave.