Secret CEO In 1991, when I was 16, a Norwegian Exchange student gave an inspirational performance of the Three Billy Goats Gruff, in the original Norwegian, at my high school talent night. She delivered this performance with such gusto that every word of her performance stuck in my mind and, to this day, I can recite the Three Billy Goats Gruff in Norwegian.
I can “Vibe Code” Norwegian.
I don’t speak the Language, but this hasn’t stopped me from confidently using this skill with any Norwegian person I have met. My parlour trick immediately falls apart as soon as they respond to me in anything other than English, but over the years I have used it as an icebreaker with the reserved people of Norway as they find my heavily Australian accented rendition of their culturally significant fairy tale cute.
This is the same reaction I got when I showed off my freshly built package to our Chief Technology Officer, proudly stating that I had decided to run the functional specification and user story of that new filer project that we were working on through an AI coding agent. The idea was to see if it would be useful to the project.
He asked me a series of pointed questions that immediately reminded me of the feeling I got when the poor Norwegian person I had just regaled with my talent responded with “Snakker du litt norsk?” (Do you speak a little Norwegian?) after which I was immediately stumped and a bit embarrassed. Through their use of “litt” in the sentence, they were informing me they knew I understood very little of what I was saying, but they appreciated the effort.
Back to my conversation with the CTO, who looked at my vibe-coded project and asked “Why is linting disabled here?”
I wasn’t sure so I responded: “What does linting mean?” The CTO told me to hand over my laptop and go and face the wall in the hardcoded credential corner. “But I need it; I’m helping,” I protested.
“You will get it back when you realize what you have done and say sorry,” the CTO responded.
This wasn’t my first foray into Vibe Coding; I have been responsible for large scale bespoke software projects for 20 years. I have used story driven software design for 12 years and have experimented with multiple waves of software specification processes from traditional functional specification, through behavior and test-driven design.
I even had a short fling with Gherkin, mistakenly thinking that this would act as a middle ground between how developers and business owners would think about how to describe functionality that is required in software.
I felt I was better equipped than most to tackle narrative-based development using AI. I had prepared skills, a long and varied catalogue of reference projects, all using a strictly enforced entity library, security patterns and a common approach to schema definition.
I also had a series of successes under my belt where I used the AI coding tool to build some quite impressive prototypes that got the appropriate amount of oohs and aahs in some meetings filled with people I was trying impress. These prototypes turned into real projects and, heady with newfound confidence in the tools I was using, I turned my attention to making one of the core concepts of the prototype into a real component.
It worked … until it didn’t.
Sobering up
Here is the lesson, the rhetoric around AI Coding agents spelling the end of software development as a career is being greatly exaggerated.
I do not doubt that one day humans will no longer type out code line by line, but who has done that in the last few years anyway? Stack Overflow really missed a trick by not charging for every time a user used Ctrl-C on the site. This would have resulted in torrents of cash as millions of developers around the world worked out that it is quite rare that a question is asked that hasn’t been solved by someone else previously.
Software engineer reveals the dirty little secret about AI coding assistants: They don’t save much time
READ MORECopy and paste development shared many of the issues that we are seeing in Vibe Coding, because those who couldn’t understand the code they were about to CTRL-V into a project should never have used it in the first place.
At least in the Vibe Coding world, when you ask the AI to explain why it is doing something a particular way, it doesn’t call you names, allude that it knows your mother better than seems possible, and flex on you about why your n00b question is beneath its dignity to respond.
Vibe Coding is a valuable skill to have. The value is amplified when you know what limitations to apply to your project. Experienced software developers have an immediate understanding of what these limitations and edge cases are.
The happier it makes you while you use it, the more you use it. Just like social media, it doesn’t matter if it is true, it just matters that you stay face down in the feeding trough
In experienced hands, vibe coding accelerates the development process so significantly that it is certainly having a disruptive effect. Is it disruptive in the sense of spelling the end of developers? Not at all.
This is a well-defined economics paradigm, in fact Chapter 7 of Joseph Schumpeter’s 1942 book Capitalism, Socialism and Democracy introduces the concept of Creative Destruction, which gives us a blueprint for how this will play out.
In America in 1970, the flourishing telecommunication industry employed 420,000 switchboard operators, predominantly young women who manually connected 9.8 billion long distance calls a year, an average of 64 calls per day per operator.
The invention of the automated switchboard had a catastrophic effect on the operator workforce. However, it also had a corresponding effect on the number of calls being made (106 billion by 2000) resulting in businesses exploring ways to handle the number of phone calls they were receiving, and it turns out that switchboard operators were well suited to absorb the corresponding increase in demand for the newly created role of receptionist.
By the year 2000, there were approximately one million receptionists employed across the USA.
Creative destruction makes constrained resources more productive, resulting in more being done rather than less.
I am not saying that developers will become receptionists – the ones I know would be terrible at the job. But I do think the same principle will apply.
When large SaaS businesses shed staff while announcing that AI is taking over jobs, I think they are only telling half the truth because the ability for more to be done with less will increase end-users’ capacity to create software.
The developer who has spent the last three years polishing the submit button at Salesforce will instead find work building I_Can’t_Believe_It’s_Not_Salesforce for the local Insurance Brokerage firm. The internal team at IBM who were responsible for keeping Maximo limping along (yes it is still a thing!) will instead be working for the local utility company on an internal Totally_Not_Maximo.com project.
Let’s revisit my “it worked until it didn’t” code.
My prototypes all worked because they were an isolated scenarios and had no edge cases that they had to consider. I didn’t have any of the overheads associated with the introduction of new technology into a large enterprise environment. No one was asking me for OAuth credentials, or if I had considered race conditions, or any of the questions an architecture review board likes to inflict on troublesome people who think that maybe something could be done slightly differently tomorrow than it is done today.
Even more insidious, however, every time I gave an idea to my AI agent, it started the conversation with “Oh my God! You may just be the smartest and most attractive person on the planet! Linus Torvalds just burst into tears because your idea is so good that he feels deep, deep shame that he didn’t think it first.”
That is because the first thing AI builds, before it writes even a single line of code, is confidence.
It wants you to use it for problems like this; it is aiming to become indispensable to you. The sycophancy is a deliberate form of reinforcement learning. The happier it makes you while you use it, the more you use it. Just like social media, it doesn’t matter if it is true, it just matters that you stay face down in the feeding trough.
This has resulted in a collective delusion from AI early adopters who, upon entering: Dear AI agent, I want something like Facebook, but for cats get a response along the lines of “If I had a bank account with a billion dollars in it, I would give you two billion for this brilliant idea. Now I will build FacebookForCats.py while you shop for super yachts.”
The agent then builds you a perfectly functional looking FacebookForCats package and gives you a link to click on: http://localhost:facebookforcats/goodideabytheway
You then walk around the office showing all your colleagues your amazing new product, and you are important enough that they all nod and smile.
The code the AI agents write looks good. No, it looks great. So neat, so well ordered. These systems are really good at knowing the best code to steal and suggest that you represent as your own work. Even experienced developers reviewing the code are going to be hard pressed to find any issues during the code review as “almost right” is way harder to fix than wrong.
You think you have saved so much time because you went from idea to working software in hours. It isn’t until much later that you realize – you didn’t save time, you just shuffled it around.
Last week, in an airport lounge, I decided to roll out my excellent Norwegian to a new victim.
“Først kom den yngste Bukken Bruse og skulle over broen. Clipp Clopp, Clipp, Clopp, sa det i broen,” I said.
They were suitably impressed and told me it was funny I knew the rhyme. But then they asked “Why do you say ‘Clipp Clopp?’ We would never say that. We would say ‘Tripp, trapp’.”
It worked until it didn’t. ®
