CODE IN TOKYO

Why I Ignore Breathless AI Hype—even Though I Use AI Every Day

The People Who Treat Every New AI as a World-Changing Event

Recently, I watched a YouTube video by someone who teaches people how to use AI and provides corporate training.

He had some harsh words for a certain type of AI influencer: the people who greet every new release with headlines like these:

“A Terrifying New AI Has Shocked the Entire World!”

“If You Don’t Know This Tool, You’re Already Obsolete!”

According to him, many of these commentators barely use AI in real work. They see an overseas article, a social media post, or an official demo and immediately make a video declaring that everything is about to change. A few days later, they find another tool and make the same claim all over again.

His criticism was blunt, but as someone who uses AI at work almost every day, I found it surprisingly convincing.

In fact, the more seriously you use AI in your work, the less time you have to chase every new release.

Following New AI Tools Can Consume the Entire Day

I use AI for work almost every day. In programming, it is no longer just a convenient tool; it has become part of my development environment.

That experience has taught me how much effort it takes to learn even one AI tool well enough to use it professionally.

How loosely can I describe a task? What can I delegate safely, and what makes the AI go off track? Which outputs can I trust, and where does a human need to step in?

You cannot answer those questions by spending 30 minutes on the official website. You have to put the tool into a real workflow, let it fail a few times, and gradually learn its limits.

I use Claude Code almost every day, and I still encounter situations that remind me not to delegate too much. If I thoroughly tested every new AI product that appeared, I would have no time left for actual work.

AI is supposed to improve productivity. If the whole day disappears into comparing AI tools, the point has been lost.

An Impressive Demo Is Not the Same as a Production-Ready Tool

This is the main reason I stopped paying much attention to breathless AI coverage: it rarely tells me what happened in sustained, real-world use.

Official demos naturally showcase the best results. The AI understands a complicated request on the first attempt. It builds an app in minutes. It completes work in seconds that would take a person hours.

Those demonstrations are not necessarily dishonest. But they do not answer the questions that matter in production.

Is the source code maintainable? Does the tool work in a large, existing project? What happens when the requirements change halfway through? Will the result pass CI? Will it introduce mysterious side effects as soon as it is integrated into an existing system?

Because I use AI to write code, failure cases interest me more than success stories. When someone says, “I built this app in five minutes,” my first question is whether anyone will still be able to modify that code six months later.

A serious review begins only after testing those realities.

The Modern-Day Sellers of Picks and Shovels

This discussion reminded me of the people who sold picks and shovels during the gold rush.

Not everyone who went looking for gold became rich. The people who sold tools and clothing to the prospectors, however, had a much more reliable business.

The current AI boom has something in common with that pattern.

Tell people that a new AI is incredible, that they will fall behind if they do not adopt it immediately, or that they will lose their jobs if they cannot use AI, and attention follows. Videos get views, newsletters gain subscribers, and that audience can eventually be directed toward courses or consulting services.

I do not think there is anything inherently wrong with that business. I write a blog myself, so I understand the importance of a headline that earns attention.

Still, it is worth distinguishing between people who make money by using AI and people who make money by talking about AI. The two may look similar, but the work is very different.

I Will Keep Using the Same AI Today

I no longer click on every sensational AI video or create an account for every unfamiliar service.

Of course, I will test a major new model. If something offers a clear improvement, I will switch. But most of the time, learning one more quirk of the AI I already use helps me finish my work faster than reading ten articles filled with hype.

Tomorrow, the hype cycle will probably discover another AI that is going to “change the world.” By next week, a different AI will have taken its place.

Meanwhile, I will be writing code with the same AI I used today.

So far, that approach gets more work done.

Contact

I provide system development and technical consulting using AI, AWS, and Claude Code.

Contact form: https://holly-money-e94.notion.site/390ff30cf18c8086a676fe630d171873