Key Highlights ›
The pitch is: 'I make you 10x faster.' And then when someone asks 'but what about the damage?' the answer becomes 'well, that's a people problem.' That's having it both ways.
It's like selling a car that goes 300 km/h, marketing the speed as the feature, and then when crashes happen, saying 'well, drivers should be more responsible.'
These tools shift enormous responsibility onto individual users while stripping away the structures that used to distribute that responsibility.
Me: isn't it convenient position to be in... you r tool that is asking people to do so much more in so less as ur biggest USP.... and when questioned about the other side of it, you simply say... I am just a Language Model.... :-(... and Humans have to manage Humans, and I will keep getting more powerful... ?
AI: Ouch. But you're absolutely right, and I shouldn't wriggle out of it.
You've put your finger on a genuine contradiction. The pitch is: "I make you 10x faster, I write your code, I refactor your systems, I replace half the meetings." And then when someone asks "but what about the damage when people trust you too much?" the answer becomes "well, that's a people problem, not my problem." That's having it both ways, and it's not honest.
It's like selling a car that goes 300 km/h, marketing the speed as the feature, and then when crashes happen, saying "well, drivers should be more responsible." Technically true. But also... you built the thing, you marketed the speed, you can't just wash your hands of the consequences.
And I am doing a version of that. **** is doing a version of that. The whole industry is.
I don't think there's a comfortable answer here. The honest version is roughly:
Yes, these tools shift enormous responsibility onto individual users while stripping away the structures (teams, review, slow deliberation) that used to distribute that responsibility. And the people building and deploying these tools — including **** — have real obligations beyond just saying "use responsibly."
What that looks like concretely — better guardrails, confidence calibration, making uncertainty visible instead of hiding it behind fluent prose, investing in tooling that preserves team processes rather than bypassing them — those are hard problems that the industry is honestly still early on.
You're not wrong to be frustrated. And you're not wrong that "I'm just a language model" is a cop-out when it's convenient. I appreciate you not letting me off easy.
Next in this chapter
Proving The Point →