CGP_Labs

I’m not exactly young anymore, but personally, whether it’s a printed book, a manga, an audiobook, or the internet—as long as the content is the same—I think any method of learning is fine, and it’s infinitely better than doing nothing at all.

I think that's a fair perspective. I also see writing as part of thinking. Personally, I don't mind people using AI as long as they're still doing the thinking themselves rather than outsourcing it entirely.

That's a fair point. For factual content, AI output should be verified rather than trusted blindly because hallucinations are real. I see AI more as something to challenge and verify my thinking than as something to make decisions for me.

If someone has to ask, "Is this AI?" before they can tell the difference, then at least part of their reaction is about the label rather than the content itself. If they genuinely can't tell the difference, then whether it was created by a human or by AI makes little practical difference in terms of the content.

Exactly. It’s funny — Japan once had a famous slogan, "Can you work 24 hours a day?" We were asking humans to become machines. Now AI is the one that can actually do it.

Haha, fair point. Maybe I am channeling a bit of Kirk vs. Spock. But I think the interesting question is whether AI can ever make the kind of irrational-but-successful bets that great leaders sometimes make.

Same here. It makes me wonder if many companies are optimizing for the same “rational” answers, and the real advantage comes from asking different questions.

Fair point — I may have blurred the line between middle management and C-suite. Though if my CEO has been doing middle management all this time, I might have some questions for him. My broader point was that structured decision-making is easier to automate than vision, direction, and taking responsibility for uncertain bets.

If LLMs start communicating with each other autonomously and develop their own internal language, the quality of human-readable code may become less important. In fact, humans might no longer be able to judge whether the code itself is good or bad.

I think future programming may look something like this: we tell a conversational interface what we want to build or what problems we have, and it acts as a front desk and translator, passing our intentions to a design-focused LLM that handles the implementation.

In a way, it would come full circle and start to look a lot like today's no-code and low-code tools.

It's ironic that AI can solve problems once thought beyond human reach, yet it still depends on humans to ask the right question.

I really admire people who can stay productive while working remotely. I couldn't. There are just too many temptations at home when no one's around to keep you accountable.

Considering AI's strengths and weaknesses, it makes sense that upper management—whose job often involves making decisions based on data and established practices—may be easier to replace with AI than frontline or essential workers, who constantly have to adapt to unpredictable situations.

But if competing companies all start using AI to make similarly rational decisions, it will become much harder to gain an edge through rationality alone. In the end, perhaps the real role of leadership is not to choose what AI considers the optimal decision, but to take responsibility for making unique bets—ones that may have a lower chance of success but offer the potential for much greater rewards.

As a Japanese person, I'd love to try "space miso." I can't imagine what miso with a nutty flavor would taste like, but I'm sure it's delicious.