Is AI Turning Coders Into Bystanders in Their Own Jobs? (msn.com) 101
AI's downside for software engineers for now seems to be a change in the quality of their work," reports the New York Times. "Some say it is becoming more routine, less thoughtful and, crucially, much faster paced... The new approach to coding at many companies has, in effect, eliminated much of the time the developer spends reflecting on his or her work."
And Amazon CEO Andy Jassy even recently told shareholders Amazon would "change the norms" for programming by how they used AI. Those changing norms have not always been eagerly embraced. Three Amazon engineers said managers had increasingly pushed them to use AI in their work over the past year. The engineers said the company had raised output goals [which affect performance reviews] and had become less forgiving about deadlines. It has even encouraged coders to gin up new AI productivity tools at an upcoming hackathon, an internal coding competition. One Amazon engineer said his team was roughly half the size it was last year, but it was expected to produce roughly the same amount of code by using AI.
Other tech companies are moving in the same direction. In a memo to employees in April, the CEO of Shopify, a company that helps entrepreneurs build and manage e-commerce websites, announced that "AI usage is now a baseline expectation" and that the company would "add AI usage questions" to performance reviews. Google recently told employees that it would soon hold a companywide hackathon in which one category would be creating AI tools that could "enhance their overall daily productivity," according to an internal announcement. Winning teams will receive $10,000.
The shift has not been all negative for workers. At Amazon and other companies, managers argue that AI can relieve employees of tedious tasks and enable them to perform more interesting work. Jassy wrote last year that the company had saved "the equivalent of 4,500 developer-years" by using AI to do the thankless work of upgrading old software... As at Microsoft, many Amazon engineers use an AI assistant that suggests lines of code. But the company has more recently rolled out AI tools that can generate large portions of a program on its own. One engineer called the tools "scarily good." The engineers said that many colleagues have been reluctant to use these new tools because they require a lot of double-checking and because the engineers want more control.
"It's more fun to write code than to read code," said Simon Willison, an AI fan who is a longtime programmer and blogger, channelling the objections of other programmers. "If you're told you have to do a code review, it's never a fun part of the job. When you're working with these tools, it's most of the job."
"This shift from writing to reading code can make engineers feel like bystanders in their own jobs," the article points out (adding "The automation of coding has special resonance for Amazon engineers, who have watched their blue-collar counterparts undergo a similar transition..."
"While there is no rush to form a union for coders at Amazon, such a move would not be unheard of. When General Motors workers went on strike in 1936 to demand recognition of their union, the United Auto Workers, it was the dreaded speedup that spurred them on."
And Amazon CEO Andy Jassy even recently told shareholders Amazon would "change the norms" for programming by how they used AI. Those changing norms have not always been eagerly embraced. Three Amazon engineers said managers had increasingly pushed them to use AI in their work over the past year. The engineers said the company had raised output goals [which affect performance reviews] and had become less forgiving about deadlines. It has even encouraged coders to gin up new AI productivity tools at an upcoming hackathon, an internal coding competition. One Amazon engineer said his team was roughly half the size it was last year, but it was expected to produce roughly the same amount of code by using AI.
Other tech companies are moving in the same direction. In a memo to employees in April, the CEO of Shopify, a company that helps entrepreneurs build and manage e-commerce websites, announced that "AI usage is now a baseline expectation" and that the company would "add AI usage questions" to performance reviews. Google recently told employees that it would soon hold a companywide hackathon in which one category would be creating AI tools that could "enhance their overall daily productivity," according to an internal announcement. Winning teams will receive $10,000.
The shift has not been all negative for workers. At Amazon and other companies, managers argue that AI can relieve employees of tedious tasks and enable them to perform more interesting work. Jassy wrote last year that the company had saved "the equivalent of 4,500 developer-years" by using AI to do the thankless work of upgrading old software... As at Microsoft, many Amazon engineers use an AI assistant that suggests lines of code. But the company has more recently rolled out AI tools that can generate large portions of a program on its own. One engineer called the tools "scarily good." The engineers said that many colleagues have been reluctant to use these new tools because they require a lot of double-checking and because the engineers want more control.
"It's more fun to write code than to read code," said Simon Willison, an AI fan who is a longtime programmer and blogger, channelling the objections of other programmers. "If you're told you have to do a code review, it's never a fun part of the job. When you're working with these tools, it's most of the job."
"This shift from writing to reading code can make engineers feel like bystanders in their own jobs," the article points out (adding "The automation of coding has special resonance for Amazon engineers, who have watched their blue-collar counterparts undergo a similar transition..."
"While there is no rush to form a union for coders at Amazon, such a move would not be unheard of. When General Motors workers went on strike in 1936 to demand recognition of their union, the United Auto Workers, it was the dreaded speedup that spurred them on."
managers argue that (Score:5, Funny)
Re: (Score:2)
"AI can relieve employees of tedious tasks" like having a job, going to work
It's not really the lack of having a job/going to work part that's a problem, it's the absence of a paycheck that really puts the proverbial monkey in the wrench. Then of course, someone will inevitably follow that train of thought to its logical derailment point and suggest that what's actually needed is to just give people money for sitting on their asses. It's an idea that sounds reasonable on the surface, but falls apart once you consider the greater economic impacts of just redistributing a bunch of
Re: managers argue that (Score:1)
"the greater economic impacts of just redistributing a bunch of money to everyone"
Is it wrong to thonk that taxation is about state violence whereas printing an indexed basic income harms no one, because you can easily index savings and private income too, as Israel demonstrsted for decades, adapting to high inflation without the broader economic impacts you handwave at?
Re:managers argue that (Score:5, Interesting)
Let's think about that for a second. There are many people who get money just for sitting on their asses. They live a good life, having done nothing of consequence. Sometimes, they've done a lot of damage. Sound familiar? I'm sure every one here can list several such people.
And yet, these people are worshipped by those who argue for the status quo. They call these people captains of industry, they call them trophy wives. They call them trust fund babies. They call them many names, but importantly, they claim that by sitting on their asses while being paid through inherited wealth and married wealth, those people are the cream of the crop that advance civilisation forward.
Something doesn't add up in your argument. Perhaps you're a victim of Crab mentality? [wikipedia.org]
Re: managers argue that (Score:1)
What if I'm able-bodied, but don't play well with others, and so no one wants to hire me? Should klercking oneself be legalized so we can cut the deadweight loss to society? Why not liberalize suicide markets?
Looking a the wrong stat (Score:2)
"How has AI affected the number of QA testers?" is more likely the question to ask.
Re: (Score:2)
Does that include people who made a bunch of money on CDOs in the early 2000's and got out before the crash? It was literally rigged gambling.
What things of value do short sellers create?
Half of finance jobs are just glorified gambling. "Something of value" was created when the stock was issued, but all the other trading is zero sum, value is moved but not created.
CEOs with golden parachutes of failing companies do not create value. The corporate raiders who bought companies to liquidate their pension funds
Re: (Score:2)
That's just not true. It could be sheer luck by placing the right bet at the right time, creating a rigged system of legal financial fraud, or maybe systematically destroying organisations by leveraged acquisitions and extracting assets. Making money at that scale often does not align with creating value in society.
Re:managers argue that (Score:5, Insightful)
Something doesn't add up in your argument.
You're conveniently ignoring two things:
#1 The extremely wealthy are a very small percentage of the population. Even if say, Musk decided to be a bigger asshole than usual and buy up the entire output of the Papa Johns pizzerias in his neighborhood, those economic effects would be limited to just his specific sphere of influence. Which brings us to:
#2 The top 1% doesn't spend their money in the same places at the same proportions, as the bottom 99%. The rich aren't buying 50 Honda Civics just because they could, or renting out multiple low-income apartments, to store their collection of the 300 iPhones that they also bought. Redistributing wealth changes how that wealth will be spent (instead of buying Twitter, it would've been spent by us "common folks" at grocery stores, on rent, car payments, and maybe a few vices too, like cigarettes, booze, and freemium gaming addictions).
Perhaps you're a victim of Crab mentality?
Far from it. I'm pointing out that simply redistributing money and hoping the free hand of the market doesn't just take it in one giant cash grab isn't a realistic solution. People need to feel they have something valuable to contribute to society, and they also need to have the means to participate in the economy. That's not something you can create artificially with a government stipend. Something like UBI is only a means of keeping the worst off members in society from starving to death, it's not a substitute for well-paying jobs that went *poof* because of automation. Telling people they'll just have to accept living off crumbs from Uncle Sam is still ultimately being dismissive of the true scope of problems that will result from job losses due to increased automation.
Re: (Score:2, Insightful)
You're also ignoring a few things
The wealthy might not be a large percentage of the population. That's irrelevant, because the amount of wealth disparity means that roughly half a dozen people control almost half the volume of currency, personally.
Maybe we don't need UBI, standard welfare would be fine... but none of it is going to work unless we reduce the functional disparity, because economies are based on moving currency. Not sitting on it and preventing anyone else from using it.
Re: managers argue that (Score:1)
"People need to feel they have something valuable to contribute to society, "
How sad is it, that you need a boss to tell you what to do?
Re: (Score:2)
They call these people captains of industry
I call this "Captain of Industry": https://store.steampowered.com... [steampowered.com]
Re: (Score:2, Insightful)
How about we don't give them money but a shelter, food, clothes and education? You know, basic necessities?
Re: managers argue that (Score:2)
Re: (Score:2)
It's an idea that sounds reasonable on the surface, but falls apart once you consider the greater economic impacts of just redistributing a bunch of money to everyone, including those who don't actually need it.
Obviously you don't realize that the enormous portions of wealth that the 0.00000000001% hold is not in the economy, it's generating passive income for people that don't want to work and whose biggest concern is who they're richer than. To give you an idea, a few years ago a certain American Royalty Rich moved their wealth from America to Europe. Their wealth was 40% of the US economy, so right now you're enjoying the economic impact of taking a whole lot of money from people who really and truly need it
Re: (Score:2)
Well, if it's in the states interest (and also corporate interest) to have people unemployed or underemployed to improve "efficiency" and "profit", then it also should be their obligation to provide those people with basic assistance (food,shelter, health care,etc.)
As we have seen over recent history corporate profits and the wealth of the ultra rich have been increasing rapidly as our economic system becomes more "efficient".
It's time to recapture some of those profits which are taken from individuals and
Just wait. (Score:2)
Every cloud, and all that.
Re: (Score:3)
"AI can relieve employees of tedious tasks" like having a job, going to work
Sadly, no, because they still need income.
To that end, I'm a little baffled by this bit:
The shift has not been all negative for workers. At Amazon and other companies, managers argue that AI can relieve employees of tedious tasks and enable them to perform more interesting work.
I would argue that's still a negative. Employees, for the most part, don't want to have less tedious work so that we have more time to do hard work for a company that will pay us exactly the same for doing it. We want less tedious work so that we have more time for the things we want to do, while still getting paid a reasonable salary.
This does the opposite of that. It takes away all the tedious stuff that provides a
Re: (Score:2)
Making a person more efficient can lead to making more money. [youtu.be]
Re: managers argue that (Score:2)
Re: (Score:2)
"AI can relieve employees of tedious tasks" like having a job, going to work
It could be worse, like having an AI as a boss.
If you have a mediocre workforce at best (Score:1)
like Amazon, then AI might be able to lend a hand.
Re: (Score:2)
Which is... fine.
How many companies really need better than mediocrity? This is a sincere question. I would argue that for most firms, hiring top tier talent is wasteful and unnecessary. You need somebody paying attention to security, for sure. But does it really matter all that much if your CRUD application suffers from insufficient inheritance, inefficient database design, or shitty code that runs fine but is just bad form?
I know this is distasteful. As somebody with three decades of background, I don't l
Re: (Score:2)
Maybe...
One of the benefits of "top" talent is influence that rubs off in other domains. 80/20 rule and all that.
On the other hand, overt specialization is suicide (which also seems to come with top talent).
The difficulty is (always) what excels in a particular environment varies dramatically. One persons' genius is anothers' prima donna.
Re: (Score:3)
I am writing this as a software developer running Claude 4 Sonnet using auto-agentic mode in another window for programming.
I can say with 100% certainty, if you are not playing close attention to what it is doing to your code it is going to do a WHOLE lot of damage when it goes off the rails. I have to constantly watch the messages scrolling by in the window as it works in order to quickly stop it before it goes off on a tangent. If you think the tools are fire and forget you don't know what you are doin
Re: (Score:2)
I just gave Claude 4 Sonnet a 30K prompt describing a code change needed in this app. I am closely watching it work to ensure it stays on topic. It will likely need to do 200-300 agentic steps to complete this request, maybe more if it introduces a lot of compiler errors.
Re: (Score:2)
15 minutes in. I had to stop it twice and get it back on the right track. So far it it has executed over 100 agentic steps. Currently it is fixing compilation errors. After it gets everything compiling it will run the new tests it made. So far It has changed 34 files to implement this change. I suspect it will take another 20 minutes for it to get the tests running.
Once the tests are running I will still need to spend an hour manually testing and reading the code before I am willing to commit it. This is
Re: (Score:2)
30 minutes. It is still trying to get things compiling. So far it has changed 3,023 lines of code. 175 agentic steps.
Re: (Score:2)
Around 37 minutes in it got lost and hallucinated. I had to stop it and explain reality to it.
At 47 minutes it got all of the tests passing. 3,062 lines of code changed. A little over 200 agentic steps.
Quick run-time test. It missed hooking the new code up to the UI so I can't trigger it. I tell it and it figures out it forgot to implement an entire screen. That will take another 15 minutes. So I will leave it running and go to bed. It is unlikely to do major damage at this stage of the process. I can see i
Re: (Score:2)
Checked back by, it's been 75 minutes. The app loads and a reasonable screen appears. It has multiple obvious bugs so I tell Claude what to fix.
So far it has touched 57 files and and modified 3,676 lines of Kotlin code with Jetpack Compose.
Re: (Score:2)
90 minutes and it is still chugging away fixing the bugs. I had to intervene two more times when it went off-the-rails.
Up to 4,415 lines of code touched.
Re: (Score:2)
Re: If you have a mediocre workforce at best (Score:2)
Your thread makes for fascinating reading. I regularly use AI for small tasks, but I've never tried anything that large.
Your description that it's like having a junior programmer at hand, who does what you tell him to - this does make me wonder how we will get junior programmers in the future. Or are the AIs advancing so fast that soon we won't even need senior programmers?
Re: (Score:2)
What turns a junior programmer into a senior on is experience dealing with problems of ever increasing complexity. Currently there is an upper bound to the complexity the model can handle; I'd estimate it at around a 25 year old programmer. It is going to take AGI to replace the senior developers.
Consider that I have been feeding the AI tasks like this for a month now on this project. I am using it to write a complete Android app. It is up to 752 files and 110,000 lines of code. Looking at my history I can
Re: (Score:2)
Re: (Score:2)
Note that I had spent an hour working with the AI to write the prompt I used at 778 lines, 29,953 bytes. And I still had to intervene a dozen times while it worked to keep it from going off on a tangent. Six months ago it wasn't even possible to do something like this.
Re: (Score:2)
this is my issue with every company i worked at. the disconnect between expectation and "virtue signaling" developers
PRs become a playground of flex where "you should refactor this", or "you could have implemented it like this", or even "why are you doing this? let's just use product X instead"
my last PR has 58 nitpicks from a dev (50 of them were style stuff, things that shouldn't even be discussed, but enforced by the linter) and the rest were valid points, product of the original requirement being incomp
Re: If you have a mediocre workforce at best (Score:2)
Re: If you have a mediocre workforce at best (Score:2)
yeah, you see, you are using phrases like "up to standards" and "less than ideal". those are not hard metrics.
those are up to interpretation and therefore "opinion". and there lies the issue. subjective comments debating whether you should abstract more here, or split this function there. those are subjective opinions that have no place in a PR. lots of "maybe we should do X because in the future we may need to do Y". to which I answer "maybe, but this feature isn't about X or Y. remember YAGNI". we'll dedi
Re: If you have a mediocre workforce at best (Score:2)
Re: (Score:2)
But does it really matter all that much if your CRUD application suffers from insufficient inheritance, inefficient database design, or shitty code that runs fine but is just bad form?
In the short run, no. In the long run, maybe. Technical debt tends to demand "interest payments" in the form of increased effort to maintain and/or debug the sub-optimal code, and it compounds over time. In the worst case, a company can end up spending its entire development budget just servicing the debt, and have no resources left over to actually accomplish any of its primary business objectives.
Re: If you have a mediocre workforce at best (Score:1)
How often do they just default and either ignore complaints or tell users tough, it's a feature not a bug?
Re: (Score:3)
And this is how you suddenly find yourself in a stagnant corporation where it's impossible to break new ground.
AIs are only decent at creating solutions from existing solutions. For new demands they'll get weird.
Re: (Score:2)
With user interfaces being reduced to "prompting" and some degenerative output for the user to base their next prompt on, most future "software" work that requires manpower will be gone soon.
How much "new" software do you actually need if you remove the user interface?
Re: (Score:2)
Evidence of AI Coding Efficacy (Score:2)
As far as evidence of AI coding efficacy goes, the NYT article cites a recent a paper by six economists [ssrn.com] - all but one current/former Microsoft Research employees - which concludes with findings that software engineers may find less than impressive: "Our preferred estimates from an instrumental variable regression suggest that usage of the coding assistant causes a 26.08% (SE: 10.3%) increase in the weekly number of completed tasks [economist-speak for weekly pull requests] for those using the tool. When we
Re: Evidence of AI Coding Efficacy (Score:2)
Re: Evidence of AI Coding Efficacy (Score:2)
Re: Evidence of AI Coding Efficacy (Score:2)
Re: (Score:3)
That's because junior SWEs are the ones writing code all the time. Senior SWEs are in meetings discussing how that code should be written.
Alternatively, it's because senior SWEs recognize that the 26% increase in task completion mostly comes from extra tasks fixing bugs introduced by junior developers who didn't take the time to fully understand the task or the code, and instead used AI to avoid having to do so. So the senior SWEs don't bother really using it meaningfully, but get very good at making it look like they are using it so that the management folks will be satisfied.
Re: (Score:2)
We are all crabs in a bucket.
The "crabs in a bucket" metaphor is more about a sabotaging of pure altruism, rather than a failure to work together for the collective good. The crabs sabotage each others' escape attempts, well, because they're crabs. But we attribute human behavioral attributes to their actions and see it as actively preventing their comrades from escaping because it provides no benefits whatsoever to the crabs who are left behind in the bucket. Presumably, their fate as someone's future seafood meal remains unchanged
Re: (Score:2)
> I think it's safe to say the human brain is not prepared for the shit storm that's coming.
I use AI daily for software development. And I have been a big proponent of it. Maybe because my job isn't at risk from AI as I co-own the business. But the business is at increased risk. As business owners and managers, we don't have an option. We have to stay ahead of this beast.
What took us years of labor to build, can now be built by AI in months if not weeks. Business leaders are feeling the pressure. In my c
Re: (Score:2)
Manus agent will be the next deepseek, from the same country.
I know what they mean (Score:5, Interesting)
Much of my coding activity these days consists of typing in prompts and reviewing the coding results which happen within a few minutes. Sometimes I have to make a tweak or two, or just tell the agent to do it. Maybe reject one approach and ask for another one. The AI will not only do what you ask, it will suggest 3-4 ways to move forward. Then I can just pick one. Definitely a different way of working.
>> One engineer called the tools "scarily good."
And I can sympathize. Apparently something that has trained on vast amounts of existing software can do a better job than you at times. Eventually it may do a better job than anyone.
Re: (Score:1)
How many "compile ai code - run and fail - find bug and rewrite" cycles do you go through?
If AI would actual compile and test their output, instead of just presenting a suggestion to review, it might actually be useful.
Re: (Score:2)
You haven't been exposed to auto-agentic mode yet. Install Augment Code in VSCode; they have a free trial. Set it to Agent mode with Auto then tell it to do something. It defaults to Chat mode. After you see what it can do, you are going to be worried for your job. Myself, I am using it to 10x my productivity.
Warning, you need to take some time planning what you ask it to do. In general it is best to first ask it to write a plan to a file and then you review that plan -- after the plan is ok then tell it to
Re: I know what they mean (Score:1)
When has the internet ever showed up in productivity stats?
Re: (Score:1)
Do you find that the output is deterministic (that is, will the same prompt reliably create the same output, no matter who the user is or what day, week, or month the prompt is asked)?
If you had to estimate a % of hallucination, what would it be?
Re: (Score:2)
It is not at all deterministic. That's because all of the tools I am using are under constant development. If everything in my environment stopped changing I suspect it would be deterministic, but that's never going to happen.
Hallucination is somewhat under your control there are three main sources 1) you exceed the context window of the AI model. For that one you just need to learn the limits of your model and not exceed them. Don't ask it to do a refactor which is going to touch two million lines of code
Re: (Score:2)
Yeah, the lack of determinism means, as you say, you cannot let by any code without reviewing, compiling, and testing.
I've found that it's fairly useful for getting a start point, but completely unreliable past that. It's also incredibly difficult to give anyone else advice on how to use it, since even if they use the exact same prompts as you did, they might get significant errors that you don't.
Re: (Score:2)
Claude 4 Opus is far better than previous models for code generation. But none of this can be used without considerable code review.
Re: (Score:2)
I'm using an IDE with AI assistance joined at the hip. $15/month subscription and well worth it. Here's my referral link;
https://windsurf.com/refer?ref... [windsurf.com]
You can run the suggested code and see if it does what you want. If you don't like it you can ask for tweaks, make them yourself, or reject it and try something different. Get the AI to write unit tests and documentation for the code if you want. And then get the IDE to commit the changes to your repository. So many tedious manual steps are now largely aut
Re: (Score:2)
I have abandoned Windsurf and switched to Augment. Windsurf is gen 1, Augment is gen 2.
I have nothing to do with either company, and i will switch again if I discover a better tool.
Re: (Score:2)
That's interesting, thanks. Looking at their website. It appears to be a VS Code extension? Which is good, I may install it.
From what I can gather they claim to have "twice the context capacity of comparable tools". You do want the AI to remember what you've been up to with it for the past multiple minutes. And then I think they say they have access to Claude 4, which I would like to see.
Re: (Score:2)
It is Claude 4 Opus the top model publicly available.
Without credit card you can get 50 prompts.
With credit card you can get 14 day free trial of the $50 600 prompt account.
Re: (Score:1)
If you were to just estimate, how many times out of 100 does the suggested code run without error?
Re: (Score:2)
I think a lot of people are doing work that consists of adding useful features to existing software or fixing bugs. Probably there already is a sizeable codebase that most developers work with. The AI looks at what you have and works from there.
It can frequently just implement a pretty close match to what you want on the first 2 or 3 passes. On occasion it won't run, usually due to editing mistakes the AI did while trying to modify its own work as per your instructions. But you can just paste the error mess
Re: (Score:1)
Actually, I have to admit that's one fairly useful feature with AIs, even when they lie - dumping a big old error log in there and having it dissect it. You can't trust it, but it sometimes finds some useful nuggets deep in the weeds that are harder to scan through with human eyes. Kinda feels like reading through stackoverflow user comments, and I trust it just about as much :)
Re: (Score:2)
I was working on a implementing a pretty difficult feature in one of my projects today and spent hours getting a solution generated with GPT-4.1, which is my current favorite. I had got it to write all the relevant code by means of a series of incremental prompts. Hit a button in a browser app, the back end spawns a shell script that runs a variety of things on the server as per your intentions and reports back. It worked pretty good but then there were edge cases where it would not quite succeed in one way
The Managers dilemma (Score:2)
Well, then (Score:1)
Hackathon (Score:3)
Product managers will program instead of engineers (Score:2)
As a former product manager you'll see more PM's start to generate code. More times than I care to count the engineer refused to hear what the product actually did and only wanted to code. I had to explain in such detail it was very similar to how AI needs prompts.
Imagine how many PM's will take advantage of getting their code created without delay. I get we aren't ready for that code to be enterprise ready. But there is no denying it, it's going to be enterprise ready soon enough.
Re: Product managers will program instead of engin (Score:2)
Re: (Score:2)
Re: Product managers will program instead of engi (Score:2)
Re: (Score:2)
In fact I did try asking this from Gemini 2.5. It did manage to create tests for the easy parts, including the corner cases, but it did not manage to create tests for the most important hard parts. It also failed to create working code for the hard parts.
But I have to admit that I used a test case for which you can not find existing source code from the internet, so AI was required to understand how physics work and generate code based on that. Nothing too hard, they teach this stuff to kids, but very uncom
Will reduce capable human programmers (Score:3)
So now we just ask the AI to either review your own code or create it based on details you feed it and voila!
That's the future most are looking at I think. People who understand just enough to tell the AI what to code but not be able to make complex programs themselves. I believe it's called vibe programming.
But I look at the flipside mentioned here for real coders. If the AI is doing all the work. I can't imagine coders remain as good at coding in the long term if reading and validating the lines is all they do during the day and don't work their brains on complex problems.
This can't be good to keep people's interest or even creativity.
Somewhere in the near future, who's going to improve the AI if good human coders become near extinct breed? Is AI going to improve itself? This can't be a good thing.
Stolen valor (Score:2, Funny)
"engineers" LMAO
Hackers paradise? (Score:3)
Given the massive quantity of unchecked code that will generate, i can see this being a massive boom for those who sell things like zero day exploits
AI can write bad code (Score:1)
I have used AI to write stuff for myself, that I have not even looked into.
I am sure it is not great, and I would likely have to spend almost as much time going thru it fixing performance and security vulnerabilities than if I had written it from the start. But it was a test. I did not write a single line myself. Used a bit of my dev skills to make it refactor code into different files etc.
So small programs for internal use will get done by AI over time. External facing is too dangerous right now.
It's like flipping burgers (Score:2)
Short order cooks used to develop a technique for it over time, but now you can learn to flip burgers by staring at a timer while various beeps and chimes are going off during your 8 hour shift. Coding will eventually be a similar sort of hellscape.
airplane pilots that become to dependent on automa (Score:2)
airplane pilots that become to dependent on automation get people killed when things go wrong.
Low quality programmers... (Score:2)
...doing simple work can be replaced by robots
Complex, novel system require work, a lot of work. AI tools can help, but the inherent nature of the problem requires expert attention and a lot of work
Code monkeys will go extinct. Experts will master the tools and accomplish much more
Like the shift from chopping your own wood ? (Score:1)
Yup, the shift from chopping my own wood to getting piped gas absolutely destroyed us
Umm.. In some ways that is true you know. But anyways i mean it super sarcastically
middle management (Score:2)
The engineers said the company had raised output goals [which affect performance reviews]
Ah, there's your problem. It's managers thinking that the productivity of programmers can be measured in lines of code or some other silly metric.
Speaking about it, that DOES explain why everyone and their dog have shifted to the bullshit "let's put every bracket on its own line" style. More lines. Clever.
So ... (Score:2)
Some say it is becoming more routine, less thoughtful and, crucially, much faster paced...
Good, fast or cheap - pick two?
(Next they'll just get another AI to be the bystander ...)
Sort of. ... Well, make that a "Yes." (Score:5, Interesting)
Disclaimer: seasoned Senior Webdev here.
Today I use AI fairly regularly for work. Meaning multiple times a week to crack difficult problems within hours or less that would otherwise take days for me to takle. It's basically a premium grade specialized Tutor/Lead Developer and an universal API documentation I can chat with. It doesn't just catch all the details in my question, answer and clarify quickly and in detail but also gives me commented example-code that I've already used in tryouts and test sessions.
After doing this for a couple of months, it is very likely that I'll book a personal coding AI subscription with Jetbrains within the next two weeks since I use their IDEs already for my daily work. Next up I'm going to let it analyze entire legacy code-bases in my responsibility and ask it specific questions about those. I expect that to go reasonably well if not really well. The trial period of Jetbrains AI offerings went reasonably well and it's plainly obvious that there is no going back when it comes to AI. The bots are here and they're taking over. ... And if you cant beat them join them. ... I guess.
My work is changing radically and rapidly and I expect this effect to grow more intense in the next 12 months. I'm preparing for what's about to roll over all of us and already started focusing on social skills 18 months or so ago and more or less abandoned learning new web technologies. That shift of focus has only intensified for me. I wouldn't be too surprised if my current job flat-out doesn't exist anymore in two years.
Bottom line: Prepare for incoming.
More like how power tools change construction (Score:3)
You used to have to hammer each nail and screw each screw. Nowadays, you use a nail gun and a power screwdriver. It makes a world of difference, and makes workers much more productive.
That's what AI does for programmers.
It's nowhere NEAR being able to work autonomously, any more than that nail gun can work by itself.
And the winner is,,,, (Score:2)
Searching for any real world examples of AI coding I came up with.... there is no such thing. What people are talking about is writing code to access large amounts of data in an AI (weighted data) environment. THAT IS NOT CODING. That is accessing a database. Again. THAT IS NOT CODING.