It amazes me how so many smart people can be so wrong. For the past 3 years many industry leaders — Sam Altman, Dario Amodei, Mustafa Suleyman to name but a few — have been claiming that in ‘6–12 months’ software developers will be out of work due to AI taking their jobs.
They see improvements in AI models and tooling and immediately jump to the conclusion that there is only one way this whole thing will end — AI will replace all software developers. It’s constantly improving so it’s only a matter of time, right?
Wrong.
Finally, someone with a bit of clout — namely Jeff Bezos — has stepped up and said it’s all nonsense. In a recent interview, Jeff said that AI won’t be taking jobs but it will in fact ‘enhance’ jobs. He stated that he thinks there will actually be a ‘labour shortage’ which is in stark contrast to the mass unemployment that others have predicted.
He went on to say that AI is going to ‘elevate’ people up to empower them to be able to do so much more. He used the analogy where up until now we have been digging the foundations of our house with a shovel and now with AI you have basically been handed a bulldozer.
From this bulldozer effect we are going to have massive productivity gains which will flood the markets with goods and services so there will in fact be deflation and the average household will be able to afford so much more.
Is he right? Who knows. These things are impossible to predict. I’m no economic expert but his thought process does make sense to me and sounds much more reasonable than the dooms day predictions coming from others.
However, if you take a closer look into the market recently, you can actually see a general change in the rhetoric from many leaders regarding AI. Sam Altman recently stating he was ‘pretty wrong’ about AI taking jobs. Dario Amodei is now realising that AI automation may expand peoples work rather than replacing them. The shift is definitely happening. It’s changing from “AI is going to replace software developers” to “AI is going to enhance software developers”. But why the subtle but definitive shift?
I have a few theories.
In 6–12 Months…
It’s been 3 years now where AI leaders have been saying ‘in 6–12 months’ all software jobs will be replaced and the markets are beginning to get tired of hearing this — It just isn’t happening. There has been no significant layoff of software developers directly linked to AI and very recently there is even an increase in the number of software jobs open.
AI Doesn’t Make You 10x More Productive
AI isn’t actually proving to make the average worker 10x more productive. There have been many reports from individuals subjectively stating that they feel 10x more productive using AI. I actually believe some of the power users of AI that this is the case. The problem is that power users are very few and far between.
For the average developer it’s just not the case and there are no tangible results in actual output from companies. Admittedly this is very hard to measure but one would expect at least a 5 times increase in release cycles, software produced and quality metrics. This just isn’t being seen in any significant way.
The Flaws With AI
They are fundamental flaws in the AI approach. AI is notoriously unpredictable. You ask the same model to do the same thing on 3 different days and the output is different every time.
Even with all this there are still massive discrepancies. I see it at work where we’re trying to scale AI usage across the team. It’s so hard to implement this well and the longer we do it the more obvious it is we will never reach 100% consistency.
AI Is Not That Great At Coding.
I use AI every day at work and it is really useful when used in a certain way. That is with clear prompts tackling specific requirements in relatively small scopes. The aim is to narrow the range of possibilities as much as possible whilst still utilising AI’s power.
As soon as you open up the scope of the requirements it starts veering off course and starts implementing things in strange ways. Not all the time but often enough that I have learnt that it pays to keep a close eye on it.
There have been no improvements with this in the last 8 months. The last significant improvement was from Opus 4.6. The newer models — Opus 4.7 and 4.8 haven’t made much difference. More recently I have just been using Sonnet 4.6 and it’s been absolutely fine.
IPOs
OpenAI and Anthropic are both gunning for IPOs. With these on the horizon they want to have less dramatic stories around their companies — it makes for a better public listing and them ultimately making more money. Therefore, they are stating that their wonderful AI tools are actually a benefit to the average human worker and not a threat like they previously predicted. It just reads nicer, makes investors happier and will generate more dollars. It would be really quite sad if this was the real reason.
Conclusion
It ultimately seems that the dooms-day mass unemployment that all the AI leaders predicted is not actually going to happen after all. Are we surprised by this? Not really but it leaves a very sour taste in the mouth that AI leaders have resorted to scare mongering tactics to promote their products.
One can’t deny that AI has and will continue to be a major disrupter. However, it seems more and more likely that AI will become part of the background just like the internet has. It will certainly play a role in our professional lives but similar to the web browser or PC, it’s used everyday but you don’t pay any particular attention to it.
Software developers will continue to have jobs to do, their value will still be evident and the world will keep turning. The likes of Sam and Dario will become insanely rich and they will live happily ever after. Just like Markus Persson.
