AI is amazing

Literally almost every day, AI technology advances and AI apps ship new tools and features.

Social media steals the show too.
The first time I heard the phrase "AI amazement merchants," I thought, "That's a sharp way to put it."
"AI amazement merchants" means people who say "This AI is amazing!" while showing tool demos that still can't survive real work — chasing buzz and followers first.

I also try anything that looks production-ready when I see it — but I still often feel it isn't there yet.
Creative AIs for banners and slides especially fail for real business work. Lots of tools are shaky on Japanese output too.

Even so, I hear AI training and AI consulting around me are making money. Smart that they've found demand — but what exactly do they teach?

Recently Ghibli-style image generation is trending. You can tell it's trained on Ghibli footage, almost certainly without permission.
Ghibli probably won't sue much anyway — people assume their legal side is weak.
Japanese business is clearly being underestimated too much.
This isn't about patriotism.
And it's not "business ethics matter" at that soft level — it's simply illegal — yet plenty of Japanese people cheer without question and shout "Illustrators are finished!"

It's "whoever does it wins," and this flow won't stop with Ghibli alone.

Nippon TV apparently holds about 30% of Ghibli and made it a subsidiary — what are they doing?
If it were Nintendo, I think the response would be faster and far more forceful.

Anyway: I use AI a lot for coding GAS, websites, LPs, and so on. I'd paid for ChatGPT for about a year, then heard "Claude's evolution is incredible!" and tried Claude.
Claude's implementation ability turned out to be genuinely revolutionary — literally at a level where you wonder if coders are no longer needed — and for the first time the AI amazement merchants' claims hit home.

At least for sites and LPs — and even MVP-level new business products including backends — you can build something that runs in a day.
I mainly use it for coding support and as a discussion partner, and it's smarter than ChatGPT.

I was so impressed I canceled ChatGPT immediately and talk mostly to Claude now.
Acquiring skills or hiring people who have them used to cost time and money — but if you can handle design and operations, generative AI can be enough.
For the first time I truly felt a breakthrough with generative AI.

After canceling, ChatGPT only answers that memory is full when I try to talk. Annoying.
People say ChatGPT Research is amazing, but it's too expensive so I haven't tried it. Besides Claude, Google NotebookLM is excellent. For free research, NotebookLM may be enough.
(I paid for Gemini Deep Research, but it wasn't impressive yet.)
DeepSeek has high technical value, I know — but I haven't tried it yet.

Demand for tech recruiting will naturally drop to some degree, and it won't take long for SaaS customers to notice they can build in-house with generative AI.
Maintenance is still needed, but the cost is far lower than hiring new people, signing SaaS, or outsourcing a whole build.
For example, more companies will decide that if they can rebuild absurdly expensive SaaS like Salesforce in-house, they should.
Things that need specialized design knowledge — accounting software and so on — are different, but even then I don't think the trend stops.
(Salesforce sales are so strong that orgs that don't really need it still buy it — layers that could get by with spreadsheets and GAS.)

Generative-AI apps have sprung up across fields these past few years too — but once Google, ChatGPT, Claude and other platforms add the feature, those apps are done.

That's about implementation. Businesses that live on information asymmetry should take damage across the board.
If you can't be on the "I make things" side, it'll be tough. The "prey on the information-weak" side will see demand fall.
In that sense I still don't think programmers who make things will disappear — and creatives and entertainment stand to gain more than they lose.
People who move their hands still win, AI or not. And if you don't keep learning tech and improving efficiency, you still drown in work-for-work's-sake.
I'm excited about becoming able to build things myself.

I always remember a Shinichi Hoshi short-short where everything runs automatically and salarymen just sit in chairs while work keeps going on its own.
I'd like to see that timeline.

That's all.