The Photography Parable, and Programming LLMs

I have recently (in the last half year) read the beautiful comparison of the invention of industrial microwaves to agentic coding - or, as the original title states: In the Future All Food will be Cooked in a Microwave and if you cant deal with that then you need to get out of the kitchen. I really recommend you read that blog here.

I have been thinking about this, and I would like to propose another comparison that may help us envision how agentic coding, or how LLMs might change the world of software development. And the world of software development is the world, as I have argued before (see for instance, here).

My suggestion is to see agentic coding in a similar way as we would regard the invention of the photographic camera.

Before the camera, creating a portrait was a complex, laborious task that almost required an education. Now anyone can create one with the right tools, and the press of a button. I'll go even further: As technology advanced, most people now carry a camera integrated into their smartphone, and can take photos whenever they feel like doing so.

Before that, only the wealthy had access to commissioning pictures - unless one was an artist or a friend of an artist. It is similar how programmers are either professional programmers, or pursuing their calling writing open source code. And similar to how programmers often get asked to help creating an app or a website.

Initially, cameras were expensive devices where only professional people had access to them. Just as how we currently have huge models that can only run on specialized and expensive hardware and their usage is limited to professional developers. Only as the technology developed, they became cheap enough to be included on everyday items until access became quasi global via smart phones.

I hope we see a similar development with LLM models, and to a degree, we already do: We have amazingly capable models (such as Gemma 4, or Qwen 3.6) which can run on moderately expensive hardware (I am running a server that costs less than USD 3k). And there are many innovations, such as the quantization aware training which can reduce hardware requirements and increase performance (see here). If these trends continue, I can see open source models becoming good enough and cheap enough for ubiquitous distribution.

Professionals might still have access to more powerful and expensive models, but that is a smaller market.

Similar to how we hire photographers for important events (such as weddings),, or for special occasions and portrait shootings; and similar to how there still are artists (painting by hand), or taking pictures with cameras. And there are many people taking pictures as a hobby. It is still a discipline, where knowing the physics, perspectives, color theory, composition, and more helps. There is still skill necessary, and a skilled artist will produce different artifacts than an amateur in many cases.

In that view, I do not see that software development will stop; rather than that software will become available to many more people who can just prompt an LLM to create a custom app / website / script for their particular use case. In a way, I see it as an opportunity to be liberated from the tech monopolies that have plagued recent times. I envision a world where bespoke software is everywhere; but where there is still space for professionals (especially where the stakes are higher, such as where money, security, health or lives are affected).

Even companies like anthropic are seeing this threat, this is why they are lobbying for more regulation and why they are likening LLMs to weapons - it is in order to protect their walled garden.

Another reason why photography is fitting as a comparison is copyright law. Initially there was a lot of resistance to awarding copyright to photos taken by cameras, but in the end, the law got adapted to the new technology.

And while I definitely do not condone the wholesale theft of intellectual property by some of the players, it is not that different from what search companies have been doing privately already for a while. And no government is going to crash that party (except for a crazy government that just wants the world to burn), simply because of the enormous amount of money involved. No one is going to want to crash this bubble; and there will be a lot of lobbying from the players.

So unfortunately I think either copyright will be adapted, either that, or it will become completely meaningless: after all, who can prove that a text was written by a human rather than an LLM?

Of course there are signs right now, but they won't remain there forever.