Introduction

A couple of weeks ago, very much against my will, I got involved with a work event trying to assess whatever the road-map for adoption of Generative Artificial Intelligence (Gen AI) is. Good lord, have mercy on my soul for that slur of corporate speech I just wrote.

How would you like yet another workspace vignette?

During this event, I had the opportunity to have a conversation with some junior colleagues as well as with one high level executive. The executive was completely clueless and her greatest contribution to the conversation was a show great enthusiasm about Gen AI. The junior colleagues were also quite clueless, but their opinions where quite illuminating.

My junior colleagues stated that they could not adopt the latest agentic coding tools provided by our employer. The key reason: they were working on the client's cloud. For the remainder of this post, please remember that these guys, among other things, write code for a living. They are professional technologists.

DISCLAIMER

Let me state that I hate agentic coding. I had one fun at work: coding. Now I do not get to code. I do not get learn stuff I find interesting. I just get to review (sometimes) mediocre code. The positive side is that, since I have nothing to prove technically, I am free to take hobby projects without any objective other than fun.

Knowledge and competence is still required

Well, I also work on my client's cloud and I get to use these agentic coding tools. They key difference between my workflow and my colleague's:

My agentic coder (let's call those a-coders because they are alpha-privative coders) generates the code locally and then deploys the code to the client's cloud service of choice.
My colleagues type their code inside the client's cloud service of choice.

In order to fully take advantage a-coders you need to know that you get to ask them to generate deploy scripts for whatever product or service your client is hellbent on using. If you cannot fathom that there is something you could do, there is no chance that a-coders will generate the required code on command.

The flip side is that while I've been adjacent to some random kubernetes clone for the past 6 months, I've managed to learn nothing about this system. Learning a proprietary configuration language is not something I am thrilled about. But what is the cost of me not learning new things? I will definitely ossify if I keep this pace. I will not get to make better decisions based on the design philosophy of this configuration language.

Overall, this point is best epitomized by the joke:

Look at the web-page I've built with the trendy a-coder. It's localhost.

Lack of agency

The more worrying note is these guys taking the world of computing as given. The whole point of writing code is: making the computer do a new thing. Why can't professional code writers think about how they could change their workflows to become more efficient?

It's not that these guys were IDE wizards that are set in their ways. They are using a web browser to edit text. And I don't mean that they code in an electron app. I mean they really code in a web browser, sometimes even without auto-completion. These people got boot-camped into technology, learnt half of every tool and made a living out of it. I am very happy for them.

But wherever I've found productivity with these tools is going out of my comfort zone. Building the things I could have never built. Even building the things I never wanted to build. We all know the schpiel, your friend tells your her great app idea, and they expect you to invest some of your time building it. I used to say no. Now I am saying: I'll build you a demo, let's talk back in 2 weeks. I do not want to build yet another cloud provider service integration in my spare time. Now I do not have to build it. The a-coder will generate it. And this is true tragedy. You can use these things to multiply your effort. To accomplish more. To have greater agency in the world. And some people are ignoring the one thing that is positive about this technology.

Executives still don't want to prompt the machine

And the next question to ask is: why doesn't your friend just do the prompting themselves? They just don't want to. I do not want either, but if I am sitting in front of the computer, I might as well prompt the a-coder in the background.

Nonetheless, these sort of idea peoples, executives and owners do not want to be associated with menial labor. They enjoy commanding people. But commanding a chat-bot is just not for them. And now I get it. I do not want to command a chat-bot either.

So there you have it. Someone with money will always be ready to pay you to do some prompting. And as such, human creativity will be reduced to an ever smaller, ever more finely 'adapted' cogs, as foretold by Nietzsche.