Ten AI Sessions for $21 a Month

I've been thinking about something that still strikes me as rather extraordinary.

My monthly AI subscription costs approximately twenty-one dollars.

On a busy day, I might have as many as ten AI conversations going simultaneously, working on different aspects of software development, research, documentation, and other projects.

And it's not unusual for me to spend ten or twelve hours a day working with AI in one capacity or another.

Now consider that for a moment.

Twenty-one dollars a month.

I've spent more than forty years developing software, and I can remember spending considerably more than that on tools that accomplished far less.

So naturally, I've wondered how this arrangement can possibly make economic sense.

But the more I've thought about it, the more I've realized that there are really two different questions involved.

What does it cost the company to provide the service?

And what is that service worth to me?

What Ten Conversations Really Means

First, I should explain what I mean by having ten AI conversations going at once.

I don't mean that I have ten AI systems continuously processing information for twelve hours.

That's not how it works.

I might be working through a particularly difficult programming problem in one conversation while another is helping investigate an unfamiliar technology.

A third might be working on documentation.

Another might be reviewing an architectural approach or helping me think through a problem that has surfaced in a different project.

Some conversations are very active. Others are waiting while I review code, build an application, run tests, or decide what needs to happen next.

I move between them as the work requires.

It's a little like having several projects spread across a desk, each with its own collection of notes, research, and work in progress.

Except now I have something capable of participating in the work itself.

That has fundamentally changed how much I can tackle in a day.

What Twenty-One Dollars Used to Buy

Think about the economics of software development.

Over the years, I've purchased compilers, development environments, libraries, components, utilities, documentation, and all sorts of specialized tools.

Some were relatively inexpensive.

Others represented substantial investments.

Many were worth every penny because they allowed me to accomplish something I couldn't otherwise do, or because they saved enough time to justify their cost.

But most tools had fairly specific purposes.

A compiler compiled code. A reporting tool produced reports. A library supplied functionality that I didn't have to write myself.

AI is different.

It can help investigate a programming problem, explain unfamiliar code, evaluate design alternatives, prepare documentation, analyze information, and develop possible implementations.

Sometimes it can help accomplish several of those things within the same project.

That's an unusually broad collection of capabilities for a single subscription.

And at approximately twenty-one dollars a month, it's difficult for me to think of another development expense that provides comparable value.

What If It Cost Two Hundred Dollars?

Here's where the economics become especially interesting to me.

Suppose the same AI capabilities cost two hundred dollars a month instead of twenty-one.

Would I still subscribe?

Absolutely.

That doesn't mean I'd be delighted to see the price increase.

But consider the alternative.

If I needed to hire people with the range of expertise I regularly draw upon through AI, what would that cost?

What would I spend obtaining help with software architecture, debugging, documentation, unfamiliar technologies, testing, and the countless other tasks that come up during development?

The expense could be substantial.

Of course, AI isn't the equivalent of hiring a team of experienced developers.

I don't want to suggest otherwise.

A good developer brings professional judgment, accountability, independent understanding, and the ability to take responsibility for a project.

AI doesn't automatically provide those things.

And there's a considerable difference between receiving a plausible answer and having a reliable solution.

But I don't need AI to replace an entire development team.

I need it to help me accomplish work that would otherwise require far more of my own time, or expertise that I might have to acquire elsewhere.

When I look at it that way, even two hundred dollars a month would be a relatively small expense.

My Time Is Still Part of the Equation

There's something else that needs to be considered.

AI isn't doing all this work while I sit back and collect finished products.

I'm still very much involved.

I'm deciding what needs to be built. I'm evaluating designs. I'm checking code. I'm testing applications. I'm identifying mistakes and determining what needs to be corrected.

Sometimes an AI makes an error that takes considerable effort to untangle.

Sometimes it proposes an unnecessarily complicated solution to a relatively simple problem.

And sometimes I have to stop a conversation because the AI has gotten so far away from the original objective that continuing would be counterproductive.

Those things happen.

After forty-plus years of software development, I have enough experience to recognize many of the warning signs.

That's part of what makes the arrangement productive.

If I simply accepted everything the AI generated, I could produce an enormous amount of software that looked impressive but wasn't necessarily trustworthy.

So when I talk about how much value I receive from a twenty-one-dollar subscription, I'm not ignoring the value of my own time and experience.

Those are essential parts of the process.

The subscription gives me additional capabilities.

My experience helps me use those capabilities effectively.

I Have No Idea What My Account Costs Them

Now let's return to the original question.

How can an AI company afford to provide all this for approximately twenty dollars a month?

I don't know.

And I don't think anyone outside the company can determine that simply by looking at my usage.

Perhaps my account is relatively expensive to serve.

Perhaps the company's infrastructure is more efficient than I imagine.

Perhaps the economics work because many subscribers use the service far less intensively than I do.

There are a number of possibilities.

The providers also have other sources of revenue, including business customers and metered developer services.

But without knowing the actual operating costs and revenue figures, drawing conclusions about whether my particular subscription is profitable would be speculation.

What I can say is that the economics look very different from my side of the transaction.

I'm paying a predictable monthly amount for access to capabilities that would have been difficult to imagine at that price only a few years ago.

And I'm putting those capabilities to work almost every day.

The Projects That Become Possible

There's another aspect of this that doesn't fit neatly into a dollars-and-cents comparison.

Some of the most valuable things AI helps me do aren't tasks I would otherwise have paid someone else to perform.

They're things I might never have attempted in the first place.

Over the years, I've had plenty of ideas for software products, tools, improvements, and other projects.

Like most developers, I've also had to make practical decisions about which of those ideas deserved the time required to pursue them.

There's only so much one person can accomplish.

And when you're responsible for designing, building, testing, documenting, and maintaining the work, the available hours become a serious limitation.

AI hasn't eliminated that limitation.

But it has changed it considerably.

I can investigate ideas more quickly. I can explore unfamiliar technical approaches without spending days or weeks getting oriented. I can work on different aspects of a project without personally writing every line of implementation code.

Sometimes that means finishing work faster.

Other times it means being able to undertake something that would previously have remained on the list of things I'd like to do someday.

How do you put a dollar value on that?

It's not simply a matter of calculating how many hours were saved.

It's the value of opportunities that become practical because the capabilities are available.

Twenty-One Dollars Is Only Half the Story

I don't know whether consumer AI subscriptions will remain this inexpensive.

There are good reasons to think the pricing models may change as the technology evolves and usage patterns become better understood.

And I certainly wouldn't want to build a business plan around the assumption that today's prices will continue indefinitely.

But I also don't want to overlook what's happening right now.

For approximately twenty-one dollars a month, I have access to an extraordinary range of technical assistance that I use throughout my working day.

I still have to supply the experience, judgment, direction, and testing.

I still have to do the work necessary to turn promising results into dependable software.

But the amount of useful work I can pursue has changed dramatically.

So when someone asks whether AI is too cheap, I understand the question.

From the provider's perspective, perhaps it is. I don't have the information needed to answer that.

From my perspective, though, the question is almost backwards.

The remarkable thing isn't that AI costs me twenty-one dollars a month.

It's how much that twenty-one dollars has changed what I can accomplish.


Related Guide: Is AI Really Too Cheap?