can you imagine a future without coding agents?
Posted by Top-Candle1296@reddit | ExperiencedDevs | View on Reddit | 45 comments
sometimes i wonder if we’re already past the point of no return with dev work. the whole ecosystem quietly shifted and now there’s this layer of agents like cosine, aider, windsurf, cursor, cody, continue dev just sitting in the background of almost every project. not because any one of them is perfect, but because the idea of building without some mix of them feels outdated.
it makes me think about the bigger picture. if this is the baseline already, what does the next decade look like?
Curious how everyone here sees it.
HappyFlames@reddit
If we look at the history of tech, we've always gone through cycles of innovation that make our jobs easier. The view that AI will replace us is a zero-sum view. Yes it may replace the current way of working but it does not mean there won't be work for us, there will just be different work. Looking at history, we basically keep adding layers upon layers of abstraction. Looking at the evolution of coding we keep find ways to write less and less boilerplate code. Imagine if we still had to write everything in 0s and 1s. Coding agents is another layer, a huge leap but I think it will create new and different work just like how web frameworks saved us tons of time twiddling with javascript quirks.
babaqewsawwwce@reddit
No one got upset when power tools were invented. Manual tool operators weren’t replaced - they got efficient.
valence_engineer@reddit
Yes they did, you've probably heard of the term that came from it:
https://en.wikipedia.org/wiki/Luddite
babaqewsawwwce@reddit
Well god damn. I’ve been educated.
ToughStreet8351@reddit
So many people deluding themselves in the comments. They are here… they work extremely well… they are going to stay. And all the people refusing to learn how to property use them and reject them altogether are simply going to be among the first to be laid off.
WrennReddit@reddit
Meanwhile, AI projects are dying off as companies realize it's dumb. Major providers brag about their AI generated code and laying off engineers because they use AI, and we have a global outage weekly.
Aicolytes keep saying they can crank out production ready projects in hours while sipping coffee. But strangely, none of them have retired from the riches they generated. They come here and bitch at engineers instead.
Methinks the MBAs are projecting on the "get left behind" bit.
ToughStreet8351@reddit
Sure… look… I can tell you my personal experience (that is also shared by most of the colleagues I talk to). It made my life so much easier… today I was able to create a script that monitored activity of a customer accessing logs, then reproduce the same requests patterns and confronted with a similar pattern using our cache and produce a report of perceived accuracy (multiple APIs and data sources to use and logs to understand) with one single prompt in less than 5 minutes during a meeting where I was active participant. That would have taken hours or more. I see the code Claude sonnet agentic mode and is better than 90% of junior and mid level devs.
Confident_Ad100@reddit
Yeah, the anti-AI sentiment here is bizarre. You would think you live in an alternative reality coming here.
I personally find it very amusing that there are so many engineers that see 0 value in AI and think it’s just going away
Sheldor5@reddit
have you ever heard of the Dunning Kruger effect?
you are on the far left of the curve ...
ToughStreet8351@reddit
A feat you might be the one too much on the far left. I agree with him and can assure you that am pretty knowledgeable in my field.
Sheldor5@reddit
AI enthusiasts are all the same, can't solve a simple problem on their own, can't think on their own, all they have is a big mouth and zero skills LOL
spookydookie@reddit
Depends on if it stays as cheap as it is. All the AI companies are bleeding money right now and they can’t afford to offer AI at these prices long term, they are just trying to get everyone hooked on it so we’ll be willing to pay $2,000/month for it and still think it’s worth it.
Confident_Ad100@reddit
Where do you all get these absurd AI costs from? OpenAI has a 4 to 1 expense to revenue, even if they 5x every plan tomorrow, they wouldn’t be more than $100.
These models are getting more and more efficient too. New GPUs and chips are more efficient, and they have found a lot of success in routing queries to smaller, more efficient LLMs if possible.
Companies like Amazon, Microsoft, Google and Alibaba are also investing massive amounts of money in AI and they are all doing great financially.
I don’t really think we are going to see AI priced out like that. The way to make money is to get people to build on top of AI tools and models, pricing out everyone would be idiotic.
Cyrrus1234@reddit
1.5 trillion in commitments for data centers are not payed by just 5x OpenAIs revenue.
Confident_Ad100@reddit
What does this stat supposed to prove?
I just have a hard time believing companies like Google will fail to exist as an AI API provider. They have the money, research, data centers, and ecosystem/users for AI in many different products.
There will be higher adoption, lower costs and potential some restructuring in the future of AI companies but they will stay around.
Cyrrus1234@reddit
I agree on Google and other big tech companies. However, OpenAI does not have a core-business besides AI. Especially AI-Datacenters suffer from depreciation, due to the specialized chips.
Meaning, besides staffing and general maintenance, you also have to regularily upgrade your chips to stay competitive. You don't just build a datacenter and pay it off for 10 years and be done with it.
You build one, constantly maintain and upgrade it (With the current pace, chips depreciate value every 2-3 years).
I'm also not arguing that they are going out of business, they won't, especially not for the established big tech companies. I'm just saying, that I don't believe a 5x price hike is enough to have a return on investment, even after 10 years.
Confident_Ad100@reddit
They are never going to raise the costs 5x, they are going to work on making the models 5-50x more efficient, which is doable and has already happened with older models (GPT4 was 10-100x more efficient after a year).
There is more that goes into a data center than just a chip. The infrastructure and the power grid by itself takes time to set up and has inherent value. And the chips itself are getting more durable and more efficient.
This is a relatively new industry, plenty of room to make things much more efficient and there is plenty of capital available to invest in figuring those things out.
I see founders working on solving these problems and creating products while this sub bitch and moans.
spookydookie@reddit
You also don’t do a 5x price hike and expect to keep 100% of your customers.
spookydookie@reddit
I think the future largely depends on if we see LLM quality improvement start to flatline and reach a ceiling like some are predicting, or if it continues to increase in quality the more money we throw at it.
If the ceiling doesn’t happen, then at a certain point LLMs will get so good that they will be likely be deemed too powerful or dangerous to allow unfettered full public access like there is today. Then you have to flip the whole model on its head and have highly restricted access only to those who can wield it “safely”. Now your whole revenue model changes.
Yeah I know that sounds a little out there and conspiracy-like, but I don’t think it’s all that unlikely at all. Democratizing power like that (if it happens) is a real threat to governments, huge corporations and other people in positions of power that currently run the show. They aren’t just going to hand over the tools to take them down.
Personally I think we’re going to see LLMs start to hit a wall and max out what we can realistically do with the current transformer approach we use today. But that doesn’t rule out another new approach that revolutionizes everything again like the transformer architecture did.
davenirline@reddit
Looks like an astroturfing post.
BoBoBearDev@reddit
No, I cannot code without a stackoverflow or the search engine/agent that finds the stackoverflow post for me.
Which-World-6533@reddit
Really...? You can't sit down and just write some code...?
No wonder candidates suck at live coding interviews.
I come from a time where you were lucky if you had documentation in a book.
Status_Quarter_9848@reddit
The question was a future without coding agents. Not whether someone can write code
Anyone with a brain can see the future clearly uses AI to write a lot of code. Choosing not to use it will eventually be like writing without an IDE now. Yeah sure you can go and write some but the use cases for it are low and decreasing. You can argue about the quality but you can always use AI then improve the quality. Just like an IDE autocorrect might speed you up but then you go and correct it.
Not such an outlandish take...
Sheldor5@reddit
I agree, at some point autocompletion is all you need
BoBoBearDev@reddit
I can code without search engine/agent just fine. But I cannot sustain it for 60sec/60min/8hr/5workday/365days/30years. There is always time that I have to use search engine/agent to help me to learn something I don't know.
Which-World-6533@reddit
You can't go 60s without referring to a search engine...? Do you have issues with your memory...?
Really...? This is something I would expect a Junior Dev to fumble on.
skeletal88@reddit
no?
why would we be? maybe you are living in your own bubble, that you think everyone is using agents and working without them is unthinkable.
i am not using any agents, i ask chathpt some questions sometimes but that is it.
it would be unthinkable if everyone who was using agents, for all work
deke28@reddit
Once the true cost is attached to them, we'll go back to doing most of the typing ourselves.
Just like COBOL... It was invented to eliminate programmers too.
BlackHolesAreHungry@reddit
The cost will go down. Once we reach the limit of LLMs which we seem to be doing they will focus on efficiency and cost
Additional-Bee1379@reddit
Saying we are reaching the limit of LLMs implies there will be no algorithmic improvements any more, which I find unlikely.
BlackHolesAreHungry@reddit
I don’t believe AGI is going to happen anytime soon
Additional-Bee1379@reddit
I didn't say it would.
codemuncher@reddit
Yes, and the emphasis of cost will be to recover as much investment as possible.
Charge what the market will bear, charge what value it delivers. If it truly can replace a dev that cost $20,000 a month to pay, it’s gonna be a lot more than $200 a month.
Confident_Ad100@reddit
It can’t replace the dev that costs $20k a month, and it doesn’t have to replace them to be successful.
spookydookie@reddit
I fully expect AI to evolve into a B2B model exclusively as it becomes financially out of reach for individual people. Replace your devs for only $7,500/month!
KatAsh_In@reddit
I feel the AI companies will charge a bomb for API based products, and keep the coding agents within affordable limits.
Main-Eagle-26@reddit
They’ll be used for niche cases.
They are simply unaffordable and the cost is not going to get cheaper before the model companies go under.
Open AI is a tinder box. It is just bleeding money with no plan for financial solubility. They’d have some kind of profitable idea by now if they were going to have one and it isn’t there.
codemuncher@reddit
Sora is a good indication of their ideas of revenue: there isn’t any.
Confident_Ad100@reddit
There are plenty of AI video generation companies that are profitable and doing 100M+ in revenue. I know at least 4.
And almost all have some sora integration.
sweaterpawsss@reddit
It seems to me that AI has already kind of reached the height of its usefulness in this present paradigm. Like, it can already write great code in most circumstances if given sufficient context. But providing “sufficient context” is still a lot of work for a human oftentimes. So the limiting factor isn’t really its ability to write code, it’s the manual effort devs still need to put in to appropriately guide it. Being able to lean on AI for architectural vision and autonomy is what would truly threaten our livelihoods…maybe we’ll get there in 10 years? I’m not the most attuned to AI developments, but I don’t see it in the current models.
Minute-Flan13@reddit
No, but I want agents to run locally. Like my compilers can. I don't want to get hooked on a service to see it enshitified or the price increase 100x.
-Nyarlabrotep-@reddit
I don't give a damn about any of them and I'll continue to code without them. I've seen the same promise repeated again and again about coding assistants or smart tools or visual design tools or whatever, and they always turn out to be overpromising, underdelivering hype. Remember how SQL was going to allow business folks to work directly with data and eliminate the need for data engineers? Or how Rational Rose was going to let software architects design systems without needing all those little programmers? This current phase is just the latest hype.
Pozeidan@reddit
Call it whatever you like, it will evolve. Right now it's coding agents, in 5 years who knows.
When the calculator was invented people didn't just stop using it after a while, the calculators just got more powerful. It's going to be the same thing. You still need someone to use the calculator.
There will be more and more coding tools and our job will gradually be to review what's been done and plan what needs to be done next.
BlackHolesAreHungry@reddit
A new programming language will emerge. Something between what we have now and a fully spec based language. We have had too many missed promises in the last decade that it is about time for it to happen anyway.
_SnackOverflow_@reddit
I don’t use coding agents on my side project. I produce higher quality code than I do at work though it take a a bit longer