That is the question.
In a world where CEO's ask us not to call AI output slop and don't be to negative, I was wondering is where is the Nobel in the tech industry?
Of course none of the current CEO's are writing their testament yet, I think.
Are they just blind for the negative consequences of the technology they created/are evolving? Platforms that are created for humans are overrun by bots. Before AI bots where only used to artificially higher or lower numbers and add predefined comments to try to fool people.
Now bots can generate content that looks human, and because it is not predefined more people are fooled.
The people that are not fooled can react angerily or don't believe it can replace juniors.
Because of AI people are thinking about creating new nuclear plants. In the meanwhile AI datacenters are running on (mobile) gas generators now.
It is not possible to build a nuclear reactor in a year, and then it also requires a some more time to start it up. It is not like a housing building where people can move in as soon as the essentials are build.
The problem I have with it that the economic force of the world are people, but we need to add solar panels to our houses to provide our own energy or depend on companies that exploit people to mass produce energy.
I applaud Anthropic makes research public that brings nuance to the AI hype. Their last research is about rich and poor country inequality.
I see people building application where AI is a part of the core. As developers we try to keep everything as performant as possible, but an AI command has a network latency and a AI generating latency. Even if the model is local there still will be the generating latency. And when models are run on servers with lower specifications the latency will go up.
Running models on less powerful servers means there will be less safeguards, while hacking the safeguards is already widespread.
If AI service providers can get the generating latency down enough it will mean vendor lock-in, and then you need to live with the price hikes of that provider.
While internet had a bubble period, when it imploded the infrastructure still existed. But if the AI bubble implodes we are left with more datacenters which have no benefit, because before AI companies were already setting up datacenters to hyperscale websites.
Like the Shakespeare quote I bastardized I have no solution at the end of this post. And because the original text was dark, I tried to match that.
While I think AI will be a tool we will keep on using, I will not be the be all end all tool that makes people unemployable in the long run.
Top comments (19)
AI can help, especially when we desperately have no clue, but we should never trust its output and always remember the principle of least power: sometimes, a full text search within a manual is still the best way to find the right answer.
There isn't really a way back I'm afraid, all we can do now is manage it and try to minimize the downsides ...
On a more positive (or is it?) note, I think we're witnessing the 4th or 5th industrial revolution, so from a financial/economic/business perspective (CEOs with $$$ signs in their eyes and dreams about surging stocks), this must be a godsend - I just hope that there will be positive "windfall" for the 'common people' as well ...
I agree there is not turning back, and I don't think it is needed.
What is the AI feature that has changed our life for good? At the moment I don't see anything world shattering like electricity or a combustion engine, or internet.
The most useful feature that I see is a productivity boost, but then you read a post like this and that shows the benefit fades fast.
I think common people are on the losing side for a while now. But that has more to do with capital concentration and political systems that only exist to sustain themselves.
Thoughtful response, especially about the common people and capital concentration etc ...
Well, when the steam engine, electricity, or the combustion engine were invented, they didn't transform the world overnight, it took time ....
Taking the example of electricity: it required a lot of ingenuity (and a lot of time) to turn the theories into practical applications - it required people like Edison, Tesla, Bell and others to come up with practical solutions, and it the took time and patience to "embed" those solutions/techniques into industry and society ...
Extrapolating that to AI - companies are now scrambling to figure out how to integrate AI into their business processes, and that's not happening as by magic, it requires time and effort to develop "best practices" ...
But I have little doubt that it's gonna be transformative, and that we're witnessing something like a 4th (5th?) industrial revolution ...
I think the comparison is more apples to oranges than apples to apples.
With electricity there are many ways to generate it, to build a model the only thing it can use is data.
Electricity is used for lighting, for heating and so on. AI can only be used on data.
AI is guided probability, and it has a place in some sectors. I don't see how every part of society is going to need it.
AI feels too limited to be as transformative than electricity.
Yes, these kind of comparisons are always a bit apples and oranges, just as the whole "nth industrial revolution" concept is (to a large extent) a bit of an artificial construct ...
Electricity is ubiquitous and is arguably THE transformative tech of the last two centuries - on that we can agree ... regarding AI, I think the way you're seeing it is too limited, but let's agree to disagree :-)
We loved your post so we shared it on social.
Keep up the great work!
Totally agree! The AI hype is real but the reality is way more complex. It’s an incredible tool, but we’re also facing some serious challenges. The infrastructure required to power AI, the energy consumption and the impact on jobs. It’s easy to get excited about efficiency, but we’re not always considering the long-term consequences.
AI can’t replace the human qualities that matter most: creativity, judgment and empathy. It’s important we don't let it overshadow the things that make us uniquely human. As we push for more AI integration, we need to think critically about how it fits into the bigger picture. Otherwise, we might end up solving one problem only to create several others.
To be used to uncover creativity and boost productivity is key, as we humans behave as managers, the AI assists us, and we assist the AI.
If long-term use causes more debt than the initial production, it is not a tool one should use, or it is being used wrongly.
More data centers can't hurt; that's decentralizing data and resources, and I see a benefit there, as long as the economics make sense, of course. Food for thought, what exactly is latency? How do we define what latency is today?
While I mostly agree with the productivity boost.
I'm more on the fence about the creativity part. While you could use it as a rubber duck, it seems like an expensive rubber duck.
If we assist AI how are we the managers? It is one of the doom thinker scenarios that AI becomes superinteligent.
I find the phrasing, the AI, a bit ominous. And there is not one AI, at the moment different models try to be the one. But as we know from software standards, when someone creates another it is just added to the list of standards.
Could you elaborate? I'm not sure what you want to express with that sentence.
While I can understand the decentralizing data argument. The problem is that the data centers they have build and want to build are enormous. That puts a strain on the resources.
Also there are not a lot of permanent jobs in a data center, so it is only a job boost for a limited time. And building data centers is not an infinite job opportunity like roadworker.
Latency is anything that slows down the responsiveness of the application. If it is in the code of your application you want that latency to be as close to zero as possible.
Of course that is an ideal scenario, that is why we try to balance latency with functionality to come to an acceptable responsiveness.
With AI the more data it has consumed the longer it takes to generate a result. That is why commands are targeting models with more or less data based on the complexity.
I think the best comparison is micro-services. When you can build the application without it is more likely to be faster.
Great piece! I really appreciate the focus on the hidden costs of AI that often get sidelined by the hype. The point about the environmental impact and the 'latency vs. performance' trade-off is something more developers should consider before making AI a core part of every application. It’s a tool, not a silver bullet.
Thought-provoking post! The shift AI is causing in various industries raises some uncomfortable truths about its long-term impact, especially in terms of economic consequences and environmental costs. The idea of AI as a tool for good, while acknowledging the vendor lock-in, inequality, and potential job displacement, is an important balance to strike. It’s clear AI has a massive role to play, but how we integrate it without creating new systemic issues is the challenge we need to face head-on.
Is like biggest companies saw a potencial on this, and they invest too much that now they can't go back. So now they are putting AI on everything. Browser with AI, notepad with AI, paint with AI, fridge with AI, doorbell with AI... And AI output is not static, so now, everything should work, but could break. So now they invest more and more on a better AI to fix what AI broke...
I want to add a slight nuance, I think the big companies saw the potential after AI specific companies got big. I think the only big IT company that is on top of AI is Google.
I think the "with AI" sticker on everything is going fase out as some point. I know a lot of people that are sick of the app for each item in your life upsell.
Before AI it was all about data collection. with AI it has some benefits but the main thing stays data collection.
What I take away from this isn’t “AI good” or “AI bad,” but that we’re rushing forward without agreeing on accountability.
A lot of the debate feels misplaced. People talk as if there’s a clean line between “developers” and “AI developers,” but there really isn’t. This is still software engineering — just at a different abstraction layer. The interface between humans and code keeps moving upward, the underlying systems don’t disappear. They just get harder to see, which is where the real risk starts.
What worries me most is not replacement, but opacity. We’re embedding AI into the core of systems while losing the ability to trace why an output happened, who influenced it, and what assumptions were baked in. Platforms get flooded with machine-generated content, and trust erodes — not because people are stupid, but because the signal-to-noise ratio collapses.
The energy point also hits hard. There’s something deeply off about framing AI as inevitable progress while externalizing the cost — environmentally and socially. We’re told individuals should optimize their consumption, while massive systems quietly scale on diesel generators and emergency infrastructure. That imbalance doesn’t feel accidental.
I also appreciate the mention of Anthropic’s research. That kind of work adds needed friction to the hype. If AI amplifies existing inequalities — between countries, companies, or individuals — then “efficiency” isn’t a neutral outcome anymore. It’s a choice.
On the technical side, the post touches something developers don’t talk about enough: latency, vendor lock-in, and dependency depth. AI isn’t just another library. It introduces network delay, compute constraints, pricing power, and security surfaces you don’t fully control. Once it’s central, walking away isn’t trivial.
I don’t read this as anti-AI. I read it as anti-amnesia.
If we treat AI as a tool, fine. But tools need provenance, limits, and memory — not just outputs. Without that, we’re not automating intelligence, we’re just accelerating complexity and hoping it behaves.
And hope is a bad systems strategy.
Thank you for your thoughtful response.
You are right it not anti AI. I do think the technology that powers AI is great. But it is also still rough around the edges.
As developers we are taught not only write happy pad tests if we want to build software that lasts.
We also need to see the negative consequences of a decision, and mitigate those.
There are too many bugs that start with the mindset of; that will never happen.
This feels like one of the few posts actually asking the hard questions instead of just repeating “AI will change everything.” Respect for the nuance here.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.