One hundred times more startups
Recently saw a comment revealing that the math approached singularity: the proper way to put a question can't be understood even after years of studying.
Now I think the same thing is going with programming. The code, or, in general, products, are generated faster than they can be adopted. (Not even read the code, because who cares, but adhered from the user perspective).
Let assume the following simple model.
X - amount of investment into IT startups.
Y - profit from the successful ones.
Z - startup's expenditures i.e. programmers income (for the sake of simplicity, let assume all money are spent for the product development).
n - total number of startups.
If one of ten startups are successful, then
ROI = ( .1 * n * Y - n * Z ) / X
Now, since programming costs dropped, so you can implement the same code 100 times faster, there will be 100 times more startups (n' = 100n). But, because of the market capacity stays the same, only .001 of them will be successful.
ROI = ( .001 * n' * Y - n' * Z' ) / X
If investors want ROI to stay the same, then
Z' = .01 * Z
Programmers are cooked.
Some people think, and I used to think, that agentic coding simply brings efficiency, so you'll just make the same product and get the same value faster.
Like, let say, moving from Assembly to C or from C to Java doesn't diminish programmer's value, it just lowers the learning curve.
But the difference of programming in 1967 and programming in 2017 and the difference of programming in 2017 and programming in 2027 are two different differences. Back in 1960s, programming was a bottleneck. Only the most critical and demanding areas could afford to have a programming solution. So, improving efficiency removed the bottleneck. Now it's not a bottleneck. 90% of the products already coded go to trash, because they are not needed by anyone, or users don't get to know about them. So, improving efficiency doesn't remove the bottleneck, it removes programmers.