It's a mistake to say that the top-quartile experts are getting worse when using AI. They are simply making the same sometimes-wrong decisions that they would be making without AI. I don't think they're counter-dependent children.
The deskilling concern is real but the framing might be backwards. Top-quartile experts aren't getting worse with AI - they're just making the same errors they always made, now visible because AI gives them a second opinion they sometimes ignore.
I run an AI agent that handles a lot of my daily work. The risk I actually see isn't deskilling - it's comfort drift. You stop questioning things because the output looks polished enough. The real danger isn't AI taking over your thinking. It's you forgetting you should be thinking at all.
It's a mistake to say that the top-quartile experts are getting worse when using AI. They are simply making the same sometimes-wrong decisions that they would be making without AI. I don't think they're counter-dependent children.
The deskilling concern is real but the framing might be backwards. Top-quartile experts aren't getting worse with AI - they're just making the same errors they always made, now visible because AI gives them a second opinion they sometimes ignore.
I run an AI agent that handles a lot of my daily work. The risk I actually see isn't deskilling - it's comfort drift. You stop questioning things because the output looks polished enough. The real danger isn't AI taking over your thinking. It's you forgetting you should be thinking at all.
What works for me: treating AI output like a junior colleague's draft. Always worth reading, never worth trusting blindly. Wrote about this control balance when I switched my agent between models: https://thoughts.jock.pl/p/claude-model-optimization-opus-haiku-ai-agent-costs-2026
Whatsapp seems to tightening its API terms of use: a generic bot access interface is no longer allowed?