AI Price Disruption: Individual Creators Benefited the Most

Hello, this is Reiji Kuga.

Since I started building my own AI workspace, I have found myself checking my OpenAI API usage costs repeatedly.

I have set the limit to 5 dollars.

It is not a very large amount. Even so, for someone like me without development experience, I was a little afraid that a bug might cause the AI to be called repeatedly, causing the charges to increase without me noticing.

Even just fixing a single feature, the interaction with the AI does not end in one go. Fix it, run it, something else breaks, and then go back. Before sending an instruction, I sometimes found myself wondering if I really needed to ask this right now.

Meanwhile, OpenAI has lowered the API price for GPT-5.6 Luna by 80%. Input is 0.20 dollars per million tokens, and output is 1.20 dollars. It has reached a level where even individuals can try and fail more easily without worrying about the costs as much as before.

At first, I just thought it had become quite cheap.

But looking back at what I am doing now, I feel like this is not just a price cut for an AI service.

What has become cheaper might not be the AI, but the cost of failure.

I went back many times to fix a single button

I am not a programmer.

I cannot write code, and even when errors appear, most of the time I do not understand what is happening just from the displayed text. Even so, while consulting with ChatGPT and Codex, I am adding features to my AI workspace little by little.

It happened the other day when I was creating a feature to duplicate saved artifacts.

The original artifact was ID 3, and the artifact created after duplication was ID 4. On the screen, there was a button to open the new artifact and a button to return to the original artifact.

However, even when I pressed them, it would not navigate to the correct artifact.

Even though the numbers seemed to change on the screen, the details from before the duplication remained in the background. I had not been able to manage the original ID and the new ID as separate states.

Check the cause, have it fixed, and run it again. This time, even if I could open the new artifact, the movement to return to the original felt off. Even if it looked like there were no problems, I could not rest easy until the tests were finished to ensure it did not affect other features.

If you only look at the finished screen, I don't think it conveys anything at all.

Press the button, and the intended screen opens.

From the user's perspective, that is all there is to it.

However, I failed many times before I could make that the norm. It is not as if AI provides the correct answer from the start, nor am I able to give the correct instructions from the beginning.

So, when I saw the news about prices dropping, my first thought was not about saving money.

With this, I can try again.

That sense of relief was greater.

What became cheaper was not the correct answer, but the trial and error.

If you use AI, a finished product comes out from a short instruction.

Looking from the outside, I think that is how it can feel. Before I started using it myself, I also imagined it would be a little simpler.

In reality, I rarely adopt the first response as is.

If I have it write an article, I break up parts that are too polished. If I create an image, I point out discomfort in the pose or composition. In development, I have to check not only the modified function but also its surroundings.

As we interact repeatedly, it feels less like the AI is giving the correct answer and more like I am gradually discovering what I was looking for.

This is similar to the time I spent making the metaverse drama 'Crossroad'.

There were times I reshot scenes that were only used for a few seconds.

Even if it worked as a video, it felt a little off when considering the characters' feelings. The positioning looked unnatural. The pause was too short, and the scene moved on before the words could land.

Even if you put the before and after shots side by side, many people might not notice the difference.

Even so, for us who are making it, it bothers us.

In the completed main story, neither the unused footage nor the time spent hesitating remains. However, I don't think the work would have taken its current form without that invisible trial and error.

What increases as AI prices drop should not just be the number of finished products.

You can shape it once, go back if you think it's wrong, and try again in a different way. It has become easier to bring ideas that would have otherwise disappeared while stuck in your head to a point where you can judge whether they are good or bad.

In creation, this number of times you can try can lead to the quality of the work.

It does not necessarily mean it will always get better.

There are days when, no matter how many times I try, I cannot surpass my initial idea. Even so, I am standing in a slightly different place than I was back when I couldn't bring anything to life for fear of the costs.

It is not simple to say that individuals benefit more than companies

If you look only at the monetary amount, companies that use AI in large quantities likely see a greater reduction effect due to price drops.

The scale of the money I spend and the amount of data companies handle are different. Within companies, there are also small teams that can try out ideas immediately and companies that move quickly.

Therefore, I cannot simply say that individual creators always benefit more than companies.

Even so, individuals have the strength of being able to try things based solely on their own judgment.

I can start building a feature I thought of in the morning that same night. If I think the angle of an article is weak, I can try out a different idea on the spot. I am the one who suffers if I fail, but I am also the one who decides whether to stop or continue.

Individual creators have limited funds and manpower.

I, too, have stopped when I had something I wanted to create because I thought I couldn't do it myself. There were times when I didn't have the budget to hire an expert, and by the time I finished researching, I ran out of time, leaving only the idea behind.

AI has not erased all of those limitations.

However, it has brought me to a place where I can fail on my own, something I couldn't even stand at the entrance of before.

If you compare only the scale of cost reduction, companies might be larger. Even so, in the sense that the range of challenges I can take on based on my own judgment has expanded, I believe that individual creators are the ones who receive the strongest benefits from this.

It is not that I have become able to hold a large budget.

Even so, the number of times I can try things within a limited budget has increased.

This difference is larger than I had imagined.

What to leave behind after becoming able to create

There are also things that concern me.

I myself have become able to produce many article ideas and image concepts in a short time. In the time it used to take to think of just one, I can now compare many ideas with different angles and compositions.

Just because of that, it does not mean I have become able to decide faster than before.

Rather, because the number of choices has increased, the time spent thinking about which ones to keep has increased.

When a somewhat interesting idea comes up, should I adopt it? Is it really what I wanted to write? Am I choosing it just because it seems easy to get a reaction?

While looking at the ideas AI generates, I sometimes find myself thinking about the parts that are inconvenient for me.

Being able to create something is not the same as wanting to leave it behind.

I can have AI generate dialogue ideas for "Crossroad." It could also organize the structure and think of alternative developments.

Even so, what the characters say in that scene, which expressions to keep, and which silences not to cut—I don't want to leave those decisions to it.

As the number of times I can use AI increases, so do the options.

It should be easier, yet there are actually more things to decide. I'm not sure yet if I can say my work has become lighter.

Still, it is better to give something a shape and be able to struggle with it than to give up without being able to create it at all.

Because the price of AI has dropped, we are now able to fail a little more often.

What to choose and what to discard beyond that point.

I feel that the cost of that, at least, has not become cheaper yet.

Recommended for you!