The Promise vs. Reality of AI in Tech
I'm genuinely excited about recent AI advancements. The tools continue improving, and we're discovering more ways to optimize and enhance our work. I've personally found a good rhythm using AI to multiply my skills rather than replace or hinder them.
However, I'm concerned about the industry's direction with these improvements. Instead of creating better developers, we're often replacing them.
Rather than building amazing accessibility tools, we're creating deep fakes and virtual companions.
Instead of developing better MVPs for *real * problems faster, we're seeing low-quality products marketed as revolutionary simply because they were built without coding experience, for no purpose other than a quick cash-grab.
Literally today on LinkedIn on my way to make the original post, I encounter yet another post in my feed from a "non-tech CEO who built this in 2 hours and it will put so many companies out of business!" 🙄
The Pushback Against Expertise
When experts and masters of their craft express concerns about these products, they're frequently dismissed as:
- "Behind the times"
- "Gatekeeping"
- "Not getting it"
But the ones saying these things are often the same people who post about "losing 4 months of work in Cursor" because they didn't know about basics like version control. Or who complain their project "became so big that Claude can't properly understand it" and now they can't continue, because they have zero underlying skill.
The Multiplayer Game Incident
What prompted this reflection was seeing a poorly executed multiplayer "game" being celebrated. While it technically ran, the quality was questionable at best. Yet when this was pointed out, critics faced pushback from the creator—someone whose credibility seemed primarily based on revenue from other AI products.
UPDATE: he's doing microtransactions in the game and people are buying it... WTH am I even doing with my life...
Quality vs. Quantity in the AI Era
It's disheartening to see professionals who dedicate their careers to building thoughtful solutions for real problems being overshadowed by substandard work that somehow finds a market. This isn't solely an AI issue:
Just look at the abundance of shovelware and shameless clones on platforms like Itch.io and the Nintendo eShop
Log out of YouTube and look at the Shorts that are popular outside your algorithm: a slew of AI-generated nonsensical garbage...
It seems that consumers increasingly prioritize cheap, easy, and quick over any kind of quality. Is that what we want as creators? Is this what we value as consumers?
Finding Balance in the AI Revolution
I struggle with this industry shift and don't have easy answers. Part of me recognizes a potential market opportunity, but my professional standards prevent me from releasing something I can't take pride in. Perhaps that's just my personal challenge to navigate.
As developers in this new AI-enhanced landscape, we face a choice:
- Race to the bottom with quick, low-quality solutions that exploit AI hype
- Use AI to enhance our existing skills and create truly valuable solutions that couldn't exist without both human expertise and AI assistance
I'm choosing the latter path. It's harder, and based on what I'm seeing probably a mistake, but ultimately I truly believe it is more fulfilling and beneficial for our industry.
Your Thoughts?
I'm curious how other developers are navigating this tension. Are you finding ways to use AI that enhance rather than replace your skills? Have you seen examples of AI being used to create genuinely innovative solutions rather than just cutting corners?
Let me know in the comments!
Full disclosure: I used some of my AI tooling to help "Improve formatting for readability" and also "Soften some of the stronger language while preserving my perspective" in this post. Because the first draft was not professional 😬
Originally posted on LinkedIn, adapted for Dev.to
Top comments (12)
:)
Short anser - no. Not finding ways where AI can enhance... :)
Longer answer - over here we do a distributed database, pretty weird one... Tried a few bots - utterly useless. I think it's just that the code is so 'untypical', that there's no chance of getting any 'enhancements' from this kind of AI... :)
Even more - (this I guess goes back to
stronger language:) - I have a sneaking suspicion that people use AI predominantly for tasks they couldn't care less about. That is - mostly when they know they don't have to 'validate' the output. 'Cos otherwise there's not much in terms of productivity gains - with AI (and validation) it will take about the same time (if not longer)... :)I think untypical code is where most of the value for companies are anyway, at least the ones doing real work on solving real problems. AI is really good at spitting out "just another web app" or "just another crud service", and yeah, there are places for those, but also, really good products that already solve those problems much better than what most AIs will spit out exist. Also, those AIs are probably spitting out some of those people's code because they were illegally trained on them.
I also tend to work on problems way more difficult than most AIs can handle, even the top models have often been unable to solve my problems, but that doesn't mean they were unhelpful, as there are things they are good at. I tend to use it more for brainstorming or checking things I could have missed or helping sort through poor documentation.
AI coding tools are generally really good at Forms Over Data and Line Of Business applications where 80% of the activity is presenting data to users, validating data or iterate over a collection/set and do stuff to it. Who really wants to write a deeply nested array.map and reduce statement by hand?
I treat AI like a junior programmer. Explain very clearly what you want, review the results with care and maybe go round the loop a couple of times when it didn't quite get the logic or syntax right, but there comes a point where it is just easier to fix the remaining code by hand rather than get the AI to try and refine it.
What really concerns me is the new generation of software engineers who won't be able to get the reps in on basic structure because there is no commercial benefit in doing so. It is cheaper to give the tasks to a senior who can get AI to churn out boiler plate code and patch the mistakes, because they would have to do that with the juniors code anyway.
I agree that treating AI as a junior that needs more instruction and hand-holding is useful, but unlike my IRL juniors, my AI doesn't seem to learn from mistakes (or retain any real lessons between context switches)... and I'm always at the whim of the next model release for it to improve. In that sense, I'd much rather take a real person who is keen to learn. Because I also agree with your last sentiment, and I've posted on this before, but we need to train up the next generation, because eventually all the existing seniors will be unable to keep doing what is needed, and then we'll have a generation of developers without a clue.
You raise an interesting point about 'where 80% of the activity is presenting data to users', I haven't thought on this much, so my stats are probably wrong, but my perception is that full stack and frontend dev make up so much of the current development landscape (maybe even 80%), that this is probably what the bulk of models are trained on, so it would make sense for them to bias that way. I'd wager that unless you are training your own models on proprietary backend code and algorithms, your AI tools will forever be stuck on the solutions you see in online courses and Leetcode and alike platforms... At least until they really start thinking for themselves.
AI as a coding assistant is great.
But when you become AI's assistant, depending on AI to do all the heavy lifting, that's where the questionable quality code comes in. The author is also unable to add to/enhance their work simply because it is not actually their own.
The big thing for me has always been that helps reduce the time I need for things, I get through boilerplate quicker, but last year I made a post about how that is causing me to forget even the most basics. There is a balance. Using AI to learn should be the goal. Not using AI to do everything for you.
Same thoughts are certainly flowing through the mind of every seasoned developer these days. I could feel the excitement and disappointment when it comes to AI several times during a day, every day.
I could see the potential, but there are no shortcuts. I had some colleagues at the university dedicated to the topic - 20+ years ago. I personally believed in a future where humans will create AI smarter than us. What this brings is fascinating and terrifying at the same time.
On the other side, there is always a bunch of people looking to make a quick buck. It does not matter which technology we are talking about, why would the AI be different? Unfortunately (in this context) the AI can already deliver real results today and here we are.
Yeah, the current wave of AI is definitely a Rollercoaster, and we get drip fed just enough "improvements" to keep the hype going. I don't think this is a bubble, I just personally think much like the crypto wave from years past, the excitement is far ahead of the practical applications. And how do companies deal with having a solution without a problem? You create problems to solve so you can monetize your "solution".
I'm realizing more and more that my issues these days are far less with AI and crypto and quantum and whatever else is coming, I'm just becoming increasingly more negative about the way companies and people will just do anything for a quick buck. I finished reading Masters of Doom last year and I'm busy with Hackers: Heroes of the Computer Revolution, and theres just something about these people who did things for the love of the craft or because they wanted to solve real problems. Of course, that doesn't pay the bills, but at what point does the goal of "making enough money to solve the problem" shift to the goal of"moar money!!!"...
has been pushed by CEOs and marketers, releasing "minimal viable products" and make their customers beta testers, advocating against test driven development or writing tests at all etc. and software that automagically does everything right, just that it didn't. The AI hype is the next excuse to lay off developers, creatives, and quality assurance.
Just as traditional JavaScript developers initially resisted popular frameworks like React, Angular, and Vue, there is now skepticism about AI in development. However, much like those libraries/frameworks gained widespread adoption despite early opposition, AI is poised to become an integral part of the development process, regardless of current sentiments. While AI may reduce the need for low performing developers, experienced developers would benefit from embracing AI as a powerful tool to enhance their productivity and capabilities.
Part of the problem I see with this point of view that is most really experienced devs are realizing that the current iteration of AI tools are not the productivity enhancement they are being sold as. That said, I do still find them useful, and I'm sure they'll get even more useful, but my whole argument is that as long as companies develop these tools for the sole purpose of making money, and as long as people use them for the sole purpose of making a quick buck, they won't reach their full potential.
Another issue is that these tools shouldn't be replacing or reducing a need for low performing developers, it should be bringing them up a notch. The industry trend of dropping these people, often without first trying to help or guide, and then not wanting to replace them with juniors, is rwlly short sighted in my opinion, because how do we train the next generation of seniors?
I'm also not sure I follow your comparison with Javascript frameworks, would you mind elaborating? They feel like fundamentally different things to me; those frameworks have always mostly been free and open source, while most companies are selling us AI, and it is extremely expensive to run decent models yourself. Also, frameworks have always tried to solve a specific problem of reducing complexity and simplifying solutions, I don't recall them ever causing companies to refuse to higher an entire generation of developers, or cause non-technical CEOS to demand more from people because a framework solved their problem over a weekend. Would love to get your thoughts on this, in case I'm misunderstanding your point.
This post really resonates with me. AI is undeniably powerful, but its true potential lies in enhancing skills rather than outright replacing them. There's a clear divide between those who use AI to augment expertise and those who chase hype-driven shortcuts.
I firmly believe AI should be a tool for creativity, not just automation. For instance, I’ve seen AI transform the digital art space—not by replacing artists, but by helping them bring ideas to life faster and more efficiently. A great example is AI Cartoon Generator, where AI empowers users to create unique cartoon-style images without requiring advanced design skills. It’s about unlocking possibilities, not devaluing craftsmanship.