I don’t know, it feels like human creativity is being converted into coding language. It all loses significance once it’s commercialized to sustain the endless cycle.
That’s my secret.
I don’t use AI.
Not a bit, quite the opposite. Before AI, I was getting weary of the repetitive nature of programming, and AI has truly helped with that. I often found myself saying, ‘If I have to write one more for loop…’. Thankfully, I write much less now, often using voice commands, too. It’s like a dream come true, allowing me to do even greater work now that I can code faster.
This generative AI wave, particularly with LLMs, is quite sneaky. Those making decisions without technical knowledge are easily fooled by this AI hype.
The anticipated inevitability of AI is overshadowed by its glaring incompetence, demanding a lot of human guidance along the way.
I’m exhausted by the excuses and justifications for this so-called AI from developers who should comprehend better.
In short, LLMs are not understanding machines; their so-called intelligence is a facade with no inherent comprehension—nothing but a pile of nonsense. You cannot rely on a Co-Pilot for areas where you lack clarity.
You must understand what you don’t know to address its bizarre outcomes. To know if a recommendation makes sense, you should know your syntax—which might be right but is something you can look up easily.
Honestly, Rust’s compiler and Clippy are far more helpful than a tool like Co-Pilot. Code automation produces reliable results, and you can’t achieve that with an LLM. Prompt-engineering becomes mere guesswork. LLMs are the least efficient way to achieve any task, as you could easily embed domain knowledge into any specialized system, achieving superior results for less effort.
Is this hype shifting coding? Yes, the lower end of the market is certainly evolving, but there’s a distinction between:
-
software engineering where you create genuine systems and components, and actually write code
-
template tweaking on WordPress which many sites still employ two decades later.
-
companies attempting to hire waves of junior programmers and wasting fortunes on trendy tactics like ‘serverless’ or ‘everything in Kubernetes’ for what’s currently in vogue (which is often a target for marketing AI ambition).
The landscape is already diverse, showing an array of experiences categorized as ‘the industry.’
Are big corporations like Microsoft going all in? Yes.
But what’s in it for you by following that example? Probably not much.
I regularly observe lost programmers asserting that AI assistants help them during initial phases, which baffles me: if you hand over system design and conceptualization to an AI unable to understand, you eliminate your role entirely, in a perfect world where easier coding and restructuring is the goal.
The genuine revolution in computing was and is about AUTOMATION; the last significant change involved infrastructure as code.
@Fallon
People looking to AI for software design might not find great results, but from my experience, LLMs offer massive efficiency gains for coding.
I’ve been testing Cursor for about a month, and honestly, I can’t go back to coding without it.
Even just the autocomplete saves me more than an hour daily that I’d spend on tricky copy-pasting or editing. I can alter the API return format and it knows to change the properties in the related React component. Suddenly, a 20-minute task can be a 5-minute job. This adds up to substantial overall efficiency gains.
When it comes to laborious tasks like creating mock data or test setups, it’s a blessing.
Getting it to generate complete components or modules can be unreliable, but it really depends on how well you provide context. The more consistently you employ design patterns in your code, the better it synthesizes new versions of them.
There have been times when it took longer to train the LLM than it would’ve taken to do it myself, but I’ve learned to recognize what it’s good at and what it’s not.
In terms of time efficiency, it’s night and day—the benefits have truly been outstanding.
I haven’t felt this significant sense of efficiency since adopting auto-formatters. I’d rather go back to manual coding than go without LLM support.
It’s not simply AI, but if someone at work suggests “Use ChatGPT for refactoring” one more time, I might lose it.
Some folks tried this to simplify code; the outcome wasn’t pretty.
I think fresh graduates could be negatively impacted by AI, claiming to have accomplished significant projects with AI assistance without truly grasping the process—this is alarming. However, if you view AI as a tool and employ it as such, it’s no different than integrating a library. Consider this: would you stick to vanilla Node.js to create a REST API or would you use Express? Similarly, would you generate every bit of code for a breakdown project with Express, or would you let AI handle the fundamental routing, controllers, SQL connections, allowing you to focus on the real deal like endpoints and business logic?
I really appreciate it. Just had a week-long coding spree testing my limits, and I was blown away. My productivity soared as I could focus more on high-level concepts while AI handled component creation.
The downside is that it sometimes feels like I might become complacent if I keep leaning on it.
@Devin
I feel the same. It has reignited my passion for coding because I can accomplish tasks quickly. I just completed a proof of concept in two days that otherwise would’ve taken me a week.
Adair said:
@Devin
I feel the same. It has reignited my passion for coding because I can accomplish tasks quickly. I just completed a proof of concept in two days that otherwise would’ve taken me a week.
Exactly!
In just two evenings, I developed three mini applications based on the Spotify API to solve my playlist management problems.
Then I transitioned my cloud MP3 player to React in another two evenings and spent the last three days building a POC for a gift-management application that I’ve had in my ideas for years.
All this with Next.js which I had never used before.
Without AI, this would have taken me MUCH longer, and I probably would have gotten frustrated with all the unfamiliarity. With AI, those headaches disappeared, letting me focus on the overall vision and code improvements.
Even if the code isn’t perfect given it was mostly rapid development, it just takes a bit longer to polish, which I genuinely enjoy compared to the former frustrations.
@Devin
“Even if the code isn’t perfect due to the focus on speed,”
I slightly disagree. The goal of using AI was to get things done for me, and the swiftness is just an added perk. Almost the same, but not entirely.
If my task takes two weeks instead of two days or even two hours, I might lose interest. That project would end up in the unfinished pile, along with countless others.
AI has changed that for me and returned the enjoyment. I’m not looking to code for the sake of it; I aim to create programs that have a purpose.
@Devin
I think it’s fine to become lazy when it comes to repetitive work.
AI allows me to focus on bigger conceptual aspects.
As long as I can check its work, it’s a massive time saver.
I use AI mainly for tedious tasks, which is most of my workload. Only the privileged few get to work with data structures and algorithms regularly. Even optimizing pipelines in data engineering is something you do just a few times—then it’s all tedious stuff again. Thanks to AI, I have more time for the more interesting aspects and can even tackle side projects.
No, it’s simply enhanced my development speed.
Its a different style of coding. Not necessarily bad, just… different
It’s really frustrating that AI is forced into almost every application.
No, it’s actually improving.
Not necessarily, it just removes a lot of the repetitive work.