The golden era of coding

There are a lot of opinions about Agentic AI coding these days, and I have one too. On one hand we have the layperson "vibe coding" which is someone with no discernible skill or experience in developing applications who just tells an AAI tool what to create. On the other hand, we have very skilled and experienced developers and other tech folk who...well...do the same thing. The process is similar, but that is where the similarities end. The code produced by the second population is of far higher quality than that of the lay people. Eventually, the AI competency will rise to the level of the skilled coders and eventually surpass it, but for now we are in the golden era of coding.

The internet abounds with stories about clueless lay people getting pwned by AIs because these people have no experience with the type of insanity that an AI can create. They also generally have no idea of the discipline's foundational knowledge and can't spot when an AI has done something very stupid. Those are the people in the articles about files being wiped out, databases being deleted, low quality AI submissions overwhelming the Apple App store gatekeepers, and general mayhem all over.

This type of chaos is to be expected when you give normies powerful tools. A reasonable comparison is race cars. If you give me a race car, I am not going to be able to get the most out of it. I don't know how to drive that fast, it's not legal anywhere anyhow, I'm kind of big so getting into the car is a pain, and it lacks a bunch of the dials and stuff I am used to. I will probably either drive straight into a wall or be passed by everyone on the road because I am too nervous to punch it. But give that race car to a trained driver and that person will wring every last second per mile out of it. They can do that because they know what to do and, more importantly, they know what not to do. Unskilled vibe coders are the first batch of people; cluelessly dawdling or screaming along without any notion something bad could happen until it does and then they blame the race car.

The current generations of developers and other tech folk are the race car driver. We know how to build stuff. We are not using the AI to do things we do not know how to do. That would be stupid because we would be unable to check its work. We are using the AI to do the donkey work we do not want to do. Or, more probably, the work that we give to junior people. SREs, devops, and other non-coding folk write a lot of code. We generally do not write code so complex that it rises to the level of an "application", but we write a lot of integration code. Integration code is the small tools that link disparate systems together, typically through API interfaces. There is a lot of reading and trial and error involved with APIs. Although there is some standardization, it's not enough. Each one has its own type of token, some with a multi-part authentication flow. Each has its own endpoints, and they're generally versioned so there is more than one endpoint for many tasks. Each returns its data in some form, typically json, but many times it is something less portable like YAML or just text. It doesn't take too long to write an API snippet but it can take a long time to read the damn docs; something the AI tools do much faster and better than humans.

Right now. This moment. This is when the productivity of the tech sector is surging. Put the horror stories of the noobs with their silly vibe coded apps away and think about the skilled and experienced tech people who are using these tools. Apps that used to take days or weeks to create take mere hours now, sometimes minutes. The SRE or devops person that had to wait for dev cycles to get something complex done, or had to spend hours weeding through shitty docs is now freed from all that. That person is an absolute baller with vibe coding. That person knows to check where the AI is storing credentials and how they are being stored. That person knows how to convert company or personal coding requirements into rules that AI will follow. That person knows you need to store certain data outside the web root. That person knows when an app needs to be multi-threaded. That person knows the best language for a task. That person knows to watch the AI "thinking" and correct it when it goes off the rails. That person just became three people.

A common lament on the forums is that those adopting AI are stupid. These people think they are the only ones who can see that this evolution will take over the traditional human developer role. These people think that when that happens, people will be out of work. These people are either truly stupid or are new to Earth, or both. Let's go back to the beginning-ish: we flipped physical switches inside computers to enable or disable them to do things we want. We then went to assembly language and then we could "write" code instead of flipping switches. We then abstracted that to relatively plain-language coding that was compiled back into assembly. We then decided that was too slow so we created a lot of powerful scripting languages that do not need compiling and grow in capabilities and breadth almost daily. Now, we offload that to AIs. In each iteration, we stopped doing something and we started doing something else. We added one more level of abstraction to the task but at the end of the day, today's code just flips switches as it has always done.

The concept of "work" is that we trade something we have for money. In tech, the thing we have is usually knowledge. To make the whole concept of "work" work, there has to be a cross between something we can get paid for and something we know. In that order. There is a tech industry because that has value to companies and they are willing to pay for expertise. We get that expertise, either from school or experience, and we have a job. Then the industry changes and no longer wants to pay for that thing we have, we need to learn new skills so we can continue to get paid. This is just basic fundamental living on Earth stuff.

The people I mentioned above who are clutching their pearls about the future of the developer role are ignoring the fact that the job they have today did not exist 10 years ago, or maybe 2 years ago depending what they do. They try to paint a picture of a static tech industry that has never changed and when one role disappears the industry will simply shrink and have less jobs. That is ridiculous. We have generally lost the roles of milkman, phone operator, powder monkeys, burlaks, and bucklemakers (all real jobs) yet the population of the workforce continues to increase steadily. That is because we now have e-sports coaches, quantum computing techs, virtual assistants (people, not AI, although there is both), drone operators, driverless car engineers, cloud computing specialists and the list goes on and on. I do believe that coding is dead. I think agentic AI has ripped the carpet out of that discipline. But that does not translate into some cataclysmic end to all jobs. It just means that we need to change the thing we have to something we can get paid for. Like we have all been doing since the dawn of money.

For the next 5 years or so, trained and experienced tech people are the golden geese. We know how to use that tool and we can wring incredible results out of it once we are freed from the drudgery of reading docs and re-reading docs and re-re-reading docs. In return, we are sharing what we know with these fledgling AIs and they will get better just as a junior person would when properly mentored. This is the golden era of coding, but it won't last for long.