Why Demand for Hand-Written Code Will Continue in the Future

LLMs seem to be effective at producing a lot of code that mostly seems to work.

Actually, I think that’s how organizations have been approaching software for years now: make it good enough, as quickly as possible, while constantly pushing the envelope on how bad “good enough” can be. No wonder LLMs seem like a huge step forward: they make “as quickly as possible” much quicker.

Software quality is (really) bad right now, because of that approach. It didn’t start with LLMs, but they’re accelerating it.

And more LLM use means more dependency on LLMs, for both organizations and developers. When a developer is reviewing tens of thousands of lines of AI-generated code per day, there’s no way to understand them all, and it’s unlikely that leadership will want to reduce that output, so the only way to proceed is to trust the LLM.

But I think quality will become so bad that eventually some organizations will realize something needs to change.

Poor code quality runs deep, having accrued over decades of questionable industry practices. To solve their quality problem, an organization will need developers who are competent enough to solve problems throughout the tech stack, not just on top of the latest framework. That means they need to deeply understand and own code. (Not “own” as in “the code belongs to me”, but rather “I deeply care and take full responsibility for everything happening in the code”, which is only possible when they have authority to do what needs to be done.)

However, it appears that heavy LLM usage causes cognitive skills to atrophy, and a lot of developers seem to be using LLMs heavily. So I think the ability to deeply understand code will become uncommon — and therefore, extra valuable — to organizations who want quality.

How many organizations that will be, I don’t know. Enough to employ all the developers who still want to code by hand? Maybe not. Only a handful of specialized companies? Who knows. I imagine it will be more common in certain domains.

But either way, we’ll need at least some developers who can code by hand at a high level of competence. If all developers lose that ability, it will be disastrous. Like the automotive industry losing all mechanics capable of rebuilding an engine.

So, some developers will continue to write code by hand, and get paid for it. For those developers, it makes sense to maximize the amount of code they’re writing by hand now, to maintain and deepen the skills that will differentiate them in the industry, and widen their competitive edge.

Perhaps most will think investing in this approach is too risky (perhaps wisely?), or think I’m exaggerating about quality. But I think others will agree, and for them, maintaining those skills is non-negotiable.

Which are you?