It’s a weird time to be a developer.
Not bad. Just weird. You open Twitter (sorry, X?), and suddenly everyone’s a “prompt engineer.” People are bu...
For further actions, you may consider blocking this person and/or reporting abuse
This is one thing that always has bugged me with AI and coding. The job isn't to write code, no one cares about the code except the developers writing it. The job is to build stuff that users use.
If the code meets all the requirements, then the user doesn't care if AI wrote it or if a human did.
The user doesn't care if AI wrote it but, 'code meets all the requirements' still leads to the intermediary concern that the code is particularly good.
This isn't work in a factory where everyone on the line is providing exactly the same contribution. Some think and experiment more, come up with new efficiencies and performance values, or even new paradigms.
I think it's a very good point of view, but I believe that at this moment we don't have an AI capable of fully programming an application on its own without losing context or generating errors in the process of adding new features. Although how the app was made is something invisible to the user, something that isn't is the presence of bugs that we need to prevent.
I never said fully programming the application, only writing the code.
There's a huge amount of stuff that needs to happen from writing code to shipping stuff to customers. AI can help with all those stages to a degree, but it still needs guidance for sure.
Love this take. I'd like to think of AI as calculators in math class. We could have one in our hands, but if we don't know what we want to compute it's useless...And teachers only allow us to use calculators after we learn the procedure by hand.
Prompting is great for scaffolding or quick prototypes. But I truly beelieve programming is still about understanding system design, testing, debugging, and edge cases. Without that, the app falls apart sooner than later.
Totally agree
AI is the new printing press. 🙃
I ship more as a SAHM with Claude in a (nap)time block than I could as a senior software dev hand-coding. And my 2yo builds games with Cline. He'll grow up with a fundamentally different (and more efficient) paradigm for turning ideas into products than we ever did.
I think you said it best yourself: "Use AI, but don’t let it erase your curiosity. Ask why, not just how.". No extremity is good, so we should not be some modern Luddites or vibe code 100% of our projects.
Every coder has to find the fragile balance him(/her)self. For every other project. That's life.
Thank you! Yes, you absolutely still need to actually learn to code, but those that aren't integrating A.I. pair programming into their workflow in some way are going to be left behind. I hear too many grizzled graybeards saying that they refuse to use A.I. to code no matter how many advancements are made. Those folks can't read the writing on the walls. Don't blindly accept revisions, don't assume you don't need to learn to program anymore, but don't just write it off either.
Honestly, those refusing AI’s help are… well, kind of dumb-dumbs(no offense 😅). Like, why choose the harder route when you could work smarter and still maintain quality? AI isn’t here to steal your job... it’s here to support you in doing it better. Nobody’s saying you shouldn’t know how to code. But rejecting AI entirely is like insisting on using a horse when you have a car in the garage. Let it assist you, just don’t hand it the keys.
Prompting is programming for monotonous work. I feel like I've said this a few hundred times by now, but AI does not help with creating something that no one has made before. If you're making something that adds value to something, you're almost certainly making something new, and AI isn't going to help you there.
I’m going to push back here, too (and then close this tab before I become an even bigger thread stalker)! Can you just open a blank repo and say, “hey AI, build me a brand-new event-driven microservice network on K8s that auto-scales like a dream”? No — not a chance! But that doesn’t mean AI is out of the picture entirely...
The same GPT models that get labeled as “bad at code” in Copilot? They’re phenomenal at surfacing ideas and blind spots even the most seasoned dev might miss after hours at the whiteboard. Give that power to someone who knows how to prompt, and suddenly you’ve got edge cases baked in, use cases you’d only catch post-release, and a whole buffet of solutions tailored to your actual scenario — all before the first line even ships!
I swear, I’m really not trying to hijack the convo 😇, but it’s wild how many folks fire up default AI and expect instant magic — then gripe when they’re left waving a plastic wand. You wouldn’t download a JDK and expect Java to spring to life on its own, right? So why treat AI any different? 🤷♀️✨
I don't think anything you said is wrong, but I'd rather spend an hour solving a problem than an hour figuring out the "optimal prompt" to get my glorified plagiarism machine to give me the result I want. Totally just my opinion though.
Again, it's my opinion, but if I have to figure out how to correctly prompt an LLM to give me my desired output, I would rather use my brain and do it myself.
Here’s the thing: you might be the top dog in your field, but eventually, the playing field flattens out — there are no more magic shortcuts to shave off time. Meanwhile, the folks actually learning AI right now? They’ll be dropping solutions into your stack twice as fast as you can even spot the opportunity. Blink and you’ll miss ‘em. 🏃♀️💨
Plus, it’s never really about chasing the elusive “perfect prompt.” It’s just input → output. That’s the whole game. Build simple systems, get simple (but powerful) results. And let’s be real — how many times have you shipped something that actually worked well in a tech stack you never bothered to learn, and didn’t care to understand? (I’ll wait.)
It’s a bit like a junior dev insisting, “Why bother with OOP? My functional code runs just fine!” Or someone swearing CI/CD is pointless because “I can just deploy from my laptop, NBD.” You can skate by, sure. But if you’re in this game for the long haul, at some point you’re gonna want to understand the basics behind modern dev — otherwise you’re just building sandcastles at high tide. 🏖️
Personally? I’d rather spend that hour figuring out why a prompt works so well in the first place. Once you get the hang of it, crafting solid prompts is basically second nature — just like picking up any new language. NLP is its own thing and, like it or not, AI is now the bedrock of modern development. 🦋
I joined the software industry at a time when AI tools like ChatGPT were becoming the new norm. In many ways, it felt like having a 24/7 mentor — helping me fix bugs, write components, and even explain concepts I never truly understood in college.
It made everything feel… easier.
Not because the work itself was easy, but because the AI made it feel that way.
Now, two years into my journey, I’ve started to ask myself some difficult questions:
🔹 Am I truly a developer?
🔹 Or just someone who’s good at asking GPT the right questions?
🔹 What would I do if AI tools suddenly disappeared tomorrow?
Some days, I catch myself pasting the same error into GPT again and again — over days, even weeks — still not fully understanding the logic behind the fix.
And ironically, even this post… was cleaned up with GPT's help.
Yes — I’m able to deliver tasks and complete projects on time.
But deep down I wonder: what about my skill development?
What am I truly learning beneath the surface?
Everyone says:
"AI is the future. Without it, there's no job in tech."
And maybe that’s true — but I also want more for myself.
I want to think independently, debug without assistance, understand the “why” behind every solution — not just the “what” that GPT gives me.
I want to be a developer with AI — not a developer because of AI.
Will I become a senior developer someday? A tech lead others look up to?
I don’t know yet. But I do know this: the growth I want won’t come from shortcuts. It will come from struggle, depth, and slowly rebuilding the habit of learning without leaning too hard on AI.
I really like the point, it might bit down grade our creativity but it's a tool for scaffold thing fast and when it broke it still your job to figure it out and tell the AI on the problem and not just dumping the code and call it the day.
I'm totally agree with the fact you must find the balance between the help of the AI with for example the boring stuff and your own mind. Personnaly I'm also using AI like a pair programmer, It help me to think and give me different pattern to solve my issue. At the end I will choose the better solution and not do just copy/paste the generated code !
In conclusion you will be always the maker but with better tools to achieve your own goals.
In my experience, prompting works great for speeding up routine tasks: scaffolding code, generating tests, even refactoring. But when it comes to debugging, scaling, or handling edge cases, actual coding skills still matter a lot.
It feels like we’ve shifted from “writing code” to “assembling outputs,” which is efficient, but also removes some of the creative thinking that made programming fun.
The best developers I’ve worked with use both: they know how to prompt well, but also when to drop into code and solve problems the old-school way. That balance seems key right now.
Oh, I’ll absolutely push back on that “AI can’t debug or scale” claim — but only if you’ve got a dev in the driver’s seat who knows what they’re doing. Quick story: just last month, I went spelunking for rogue zombie threads that kept nuking our LLC instance. Production? Not even a peep. But this thing’s been haunting us for years — and I’ve watched everyone from SEs to Staff-level folks go toe-to-toe with it, sometimes pushing a "fix" but never really getting to the root.
Trace logs at the JVM layer with Spring ORMs? Pure spaghetti. (If you’ve never looked at one of those logs with Hikari + JDBC in the mix... don’t put it on your bucket list! 😵💫) I once left it running overnight by accident — got a love letter from the on-prem team the next morning because my “light” scheduled tasks chewed through 17 million lines of logs. Oops! 😅
But here’s the thing: OpenAI’s o-series models (even the tiny ones!) are pattern-finding monsters. Give them a focused log dump and the right context — I tossed in 3 sets of 5 minute runs of raw logs, some matching data dumps, and a config directory tour, then said, “Find the top 5 possible causes for rogue zombie threads in this app.” (Yes, my actual prompt was the length of a novella. Worth it!)
Did I get a magic answer? No, of course not! But I did solve in three days — with AI — what a parade of humans hadn’t cracked in years.
TL;DR: It’s not about if you should use AI — spoiler: you definitely should. It’s a tool, and the real magic is in how you wield it. Mix in some curiosity, a bit of training, and the stubbornness to chase answers, and suddenly you’re moving mountains. If your bookshelf’s still wonky, don’t toss shade at the wrench. 🛠️✨
Absolutely, your example highlights exactly how AI can serve as a force multiplier rather than a replacement. Dealing with complex logs, especially with Hikari + JDBC and Spring ORMs, can be overwhelming for humans alone. Providing AI with focused context, as you did, demonstrates how it can surface patterns and potential causes much faster than traditional approaches.
I also appreciate your point that AI is not a magic solution — the real value comes from combining human expertise with AI’s analytical capabilities. It’s a powerful reminder that success depends on how effectively the tool is used.
I’d be very interested to learn more about your approach to structuring prompts and handling complex logs.
"you’ll need to actually understand what’s going on under the hood." That's the most critical point. Without really understanding what's going on, it’s easy to build something fragile that breaks in edge cases. Genuinely trying to understand what is going under the hood is what makes the difference between a user and a developer.
You know, maybe I’ve been using AI wrong this whole time (but I seriously doubt it 😂). Sure, prompting is fresh for some folks, but c’mon... Just because I’m handing off the same prompt to both ChatGPT and Copilot doesn’t mean I get to skip the deep dives through Liquibase docs, or those hours spent reverse-engineering how to wrangle a solution from the mess.
If anything, orchestrating from a systems and design angle is the part I’ve always loved most — code is just the vehicle, not the destination. So yeah, dev roles are morphing. But trust me, I’m never just “editing” what AI spits out. Worst case? I’m actively steering: pushing the model for edge cases, spotting bottlenecks before the first commit, and troubleshooting stuff AI couldn’t possibly guess from context alone.
Sure, knowing JavaScript, Java, Python, Go — pick your flavor — still helps if you’re in the business of translating between humans and CPUs. But honestly? I’d rather let AI do the code gymnastics while I get busy untangling the actual problems that need solving. 🧩✨
Oh, and let’s be real: AI does not magically write production-quality, secure code by default. It takes good prompt design, system know-how, and more configuration than anyone admits in public. Can it reach that level? 100%! But you’ve gotta guide it — AI’s not mind-reading (yet). 😉
You said that you miss reading the docs and zoning out with your code editor. But you use artificial intelligence to solve your problems. Why don't you just stop prompting and go back to do what you miss? Nobody is forcing you to use artificial intelligence to write code.
I appreciate your point! AI is a tool to speed up the routine so I can spend more time on creative, challenging parts I love. It’s not about stopping prompting but balancing it with the deep, focused work that keeps programming meaningful. Using AI wisely helps me work smarter without losing that connection to the craft.
for real 💯 reading this was like a breath of fresh air. like, it is wild how fast everything’s changing tbh but knowing the why (and how) definitely feels like it could elevate everything we build, not just in making better work, but also giving us the flexibility to adapt to whatever comes next.
Great article - I severely doubt wether AI could write something like this!
I have mixed feelings about this. Here me out.
It is arguable to say "when you start writing prompts instead of functions, you stop flexing those problem solving muscles that got you into programming in the first place". It completely ignores the way we actually build software. It's more than just coding and writing functions. You still have to find the right problems to solve, architect/design solutions, test, and deliver value to end-users. So, we have no shortage of levers to keep us engaged in problem-solving.
I have always spent a great deal of time in architecture and design BEFORE even touching a line of code. I've always planned for how the code I will write will be tested and deployed too BEFORE writing a line of code. While AI is doing a bit of that too, it's making a bunch of decisions that I would NOT make in those areas. So, keeping with the idea of problem-solving, I am still doing a lot of it to deliver value...even though the coding portion has become easier.
We might be becoming less of coders, but that doesn't mean we'll ever stop being problem-solvers.
most intelligent people are not on X
Really makes you think about how AI is changing every field! At Accio AI, we see the same shift in procurement - buyers used to search manually for suppliers, now they just describe what they need in plain English and our AI handles the rest.
The key insight? Like you said about coding: AI doesn't replace expertise, it amplifies it. Our smartest buyers use AI suggestions as starting points, then apply their own negotiation and quality assessment skills.
Fascinating how this 'AI-as-copilot' model is becoming universal!
Hey! I actually code cool stuff in assembly! And I'm definitely not cursed.
AI seems to be a great tool, and we got to embrace it - without making it our entire personality. One pattern I notice within devs that use AI (or in myself when using AI), is that you just care less about the code. It's not some sacred art anymore that you stare for hours. It just becomes yet another product. You tend to test less, or skip anything that might take some time... because now your brain is programmed to be lazy.
Agree on the creativity part. If you look at all AI generated stuff, they all look the same. There's literally no flavour that must come from within you.
I did a blog post recently addressing the same issue. If yall wish show me some love, I'm trying to grow my platforms.
Thank you! And love your take on this
Thanks for sharing @seniru
Gladly – Engineering is not becoming prompting.
This is well articulated, I told my collegues we need to understand the WHY behind each line of code.
Thanks @bikodes
This nails it.....prompting is powerful but coding is still where the fun is ..
We don't need code anymore. AI does it all.
Programming has always been prompting.
Prompting has become an indispensable tool for all of us. It has augmented our productivity.
Very well explained. Thanks a lot!
Keep reading the docs and thinking about edge cases. AI doesn't understand anything, and it makes things up.
Also AI doesn't write secure code.
Yes prompting is a means to an end. But I don't think it changes programming. We have to still come up with the specifications to let AI generate the code.
Maybe it changes for developers that hate to write documentation, because that is what specifications are. You can let AI create documentation for different target audiences.
I don't use an editor with a AI prompt module, because I tend to switch AI's constantly.
I use it when I'm unsure about things. And I use it to generate code I know I can't write that fast.
Thanks for sharing your thoughts @xwero . I agree that relying solely on AI can be risky, especially for security and deep understanding. That’s why I see AI as a helpful tool, not a replacement for critical thinking or reading docs. Prompting speeds up some parts, but the real skill is knowing when to dive deeper and question the code AI generates. Balancing both is key.
"It’s a weird time to be a developer." So true, now even some doctors and marketers think they can program anything until things break. Copy and paste code from AI is one thing but understanding why it works and how to debug is another.
what i'm doing now is to save the post. Seriously. It's my first time reading something enjoyable about the 'vibe coding'. I'm good at it, but it doesn't mean i should abandon my 30min / week on leetcode
I keep getting told this from clients and others... "you better start looking for another job"... the truth is my "job" is needed now more than ever. You know who should be worried about AI? The Creatives.... I don't need an artist to make me an icon anymore, and with the background in logical thinking and problem solving skills I have honed over the years I can actually make work the crap my clients create with AI that looks good but has no substance.
Such a great read ! Really enjoyed it