For a large company, AI coding can become a huge software bill.
Thousands of employees. Premium models. Long-running agents. Token usage nobody really tracks. Finance teams trying to understand why the cost keeps going up and whether the company is actually shipping more because of it.
That problem is real.
But from my side, the math is very different.
I have a full-time job.
My projects happen before work, after work, late at night, and on weekends.
So when I use AI coding tools, I am not trying to make a large engineering organization slightly more productive.
I am trying to create capacity I do not have.
A feature in @GetLimpio that would normally take me several nights can sometimes be planned, implemented, tested, and improved much faster.
A product decision in @AddressHub can move forward without waiting until I have budget for another engineer.
A bug, a refactor, an onboarding improvement, a test suite, a product flow, all of those become more realistic when I can work with AI as part of the build process.
Without AI, a lot of what I have built in the last two years simply would not exist.
That is the real difference.
For a large company, AI can be marginal productivity at massive scale.
For a bootstrapped founder, AI can be missing capacity.
The other important part is model discipline.
Not every task needs the most expensive frontier model.
I still use stronger models when the problem needs deeper reasoning, but for a lot of implementation work, open coding models through tools like @OpenCode Go are already good enough.
And sometimes “good enough at a much lower cost” is exactly what makes the ROI work.
So I understand why companies are starting to worry about AI costs.
But I do not see AI coding as a cost problem by default.
Used without control, it becomes one.
Used with discipline, especially by a founder with limited time and limited capital, it can still be one of the highest-ROI tools available.
Same technology.
Very different math.