in AI, Software

Code isn’t cheap after all

Code was said to have become cheap

Someone wrote last year, ” Code is cheap; show me the talk”. Now the hacktoberfest (hacking incidents due to LLMs in october 2026) is going on ,it is making us question that very statement.

to be fair, Nadh wanted to make a point that writing code itself is less of a cost item now since the bottleneck of paying humans to encode requirements into programs is removed. Thus the question has shifted to how thoughtful you have been about what functionality your code serves. To that effect, it was a fair point he made.

Why software exists

This brings us to the main point that writing code or software isn’t an end into itself, as so many in our community believe it to be. Code or software has to have a real world utility!

Part of that is code being of some functional use, being performant, easy, and all the blah we all hear in design reviews. But the most important is that it has to be economical to operate! Security, you see, is a byproduct of this need.

We can, for example, have a static HTML site which can be broken each day like a rangoli in front of indian/Asian household, only be me easily remade every day. That kind of hacking or security incidents we don’t bother about much.

The one that we care about is the one that affects business continuity, service guarantees, legal liability and reputation risk! This might look like unnecessary use of fancy terms used to define security. But the buildup of NFRs on software has shaped up this way over the last many decades.

Now that AI has made coding cheap, we need to also analyze all claims of it being “cheap” in that manner. That is

Cheap for whom?

Under what circumstances? and

to what effect?

Back to basics on why software is needed

In the example we took i.e security, we have real evidence that such questions weren’t asked before announcing conclusions. And now we see the first order effect of it when different kinds of incidents are playing out.


Do we need to aks them for other aspects like performance, usability, compliance, maintainability, supply chain, pricing, obsolescence, and so on?


I say yes, in the end…wait…in the beginning, the need, the mother necessity of software is business that functions.

The necessity isn’t that AI labs have come up with models that can demonstrate that humongous amounts of working code can be generated by them. It can,there is no doubt!

But unless this ability also connects back to the real-world environment in which the business and the software it made/bought/linceced operates and be liable for, this claim is a pure CTO-class claim not fit for consumption by other CXO roles¹.

AI is making us go back to the reasoning of why software was needed in the first place (but we went into why developers are needed in the first place).

PS: which is also to say that the development lifecycle of code is different than operating software! AI labs have an answer for the first (as of now )

____

Footnotes :

¹ At the heart of this distinction lies the belief that software and automation is some sort of inevitable shift upon humanity. But we see that while digital payments help at some scale, mobility and behaviour design, cash and tokens are many times more effective and as reliable. Second, we need to also be aware that an abundance of supply doesn’t always generate demand, let alone justify it. For example, Mobile phones practically can do all the work you can imagine in your daily life without the need for human touch; drones can drop kids into schools; one can survive just by eating vitamin pills and protein powder; just their existence doesn’t create demand , there has to be real need for it or the habit has to be enginnered !


Write a Comment

Comment