I appreciate your honestly but you aren't going to find hardly anyone on Lemmy who wants to read anything made by generative AI. People have a strong negative reaction to it. As someone who tried to give AI an honest chance, I found it less than useless for even the most basic tasks. It often has no idea what I'm talking about or when it does it just makes shit up. It's good for laughs though, I had it write a script for a sequel to the 2008 film Speed Racer. That was funny to read lol.

[–] dependencyinjection@discuss.tchncs.de -5 points 6 days ago (1 children)

Have to ask what you used it for for it to be “less than useless” as that seems mighty biased or hyperbolic.

I’m a software developer and work for a tiny company of 7 employees including the owner engineers.

We use LLMs everyday and the time it has saved us in real world numbers is massive to say the least.

It was enough that we all got pay rises and a reduction of hours from 37.5 to 33 per week.

It ain’t doing our jobs for us but to say it is useless is ridiculous and shows the clear bias of Lemmy on this subject matter. It’s a kin to saying a calculator is useless in making an accountable faster at their job.

I like this place but at this point it is nothing more than an echo chamber akin to the Donald where discussion is not encouraged for dissenting opinions and it’s just a circle jerk of everybody feeling superior without any intention of making the world a better place.

[–] kopasz7@sh.itjust.works 0 points 6 days ago (1 children)

Ironic user name for a vibe coder. Do you guys have an estimate for the rate of technical debt and vulnerabilities introduced or is that secondary to "getting it done"?

[–] dependencyinjection@discuss.tchncs.de 0 points 6 days ago* (last edited 6 days ago) (1 children)

Vulnerabilities is going to be the same as before as it’s not like we just vibe code the whole thing. It’s akin to Intellisense on steroids.

As an example we have pretty well defined coding standards and so there is a lot of repetition in how we make calls from Typescript -> GraphQL -> C#. And the fact that CoPilot learns from our code base and then learns how we do thing.

Why do I want to write the same mutations or queries every time when it’s the same, when LLMs can do it and we can solve NEW problems not the same one over and over.

Technical debt is certainly down, because take me as recently promoted from Junior to Dev and my bosses hourly consulting rate would be £1000-2000 an hour then me asking less questions saves him a lot of time to actually do the higher level stuff.

I find it interesting how you would question that my boss, literally the smartest person I’ve ever met in my life wouldn’t have done risk assessments and made a value judgement. After all we get paid more now for less work.

We don’t just get things done and we go above and beyond for clients. If we quote £50k for an application and we expect it to take 6 months and the client is still refining 12 months later then we do what is right by the client and don’t charge more.

We have local councils, one of the largest phone providers in the Uk, and other massive clients that always come back for repeat business and recommending us to other people. We do not advertise and we are not short of work.

I have to wonder how many of these people saying LLMs are useless in this context actually earn money coding and how many just rant about it because it’s the trend here.

[–] kopasz7@sh.itjust.works 1 points 5 days ago

Vulnerabilities is going to be the same as before

Because... assumptions?

Why do I want to write the same mutations or queries every time when it’s the same, when LLMs can do it and we can solve NEW problems not the same one over and over.

Code reuse instead of code duplication?

Technical debt is certainly down, because take me as recently promoted from Junior to Dev and my bosses hourly consulting rate would be £1000-2000 an hour then me asking less questions saves him a lot of time to actually do the higher level stuff.

'Technical debt' is not akin to financial debt. More AI generated code will result in more code that nobody wrote, thus nobody knows. Over time, people come and go, but the code stays with all the accumulated cruft. The spaghetti.

I find it interesting how you would question that my boss, literally the smartest person I’ve ever met in my life wouldn’t have done risk assessments and made a value judgement.

Sound like you are a junior dev, so how many people have you worked with for this to be a generally meaningful statement? It's also appeal to authority.

After all we get paid more now for less work.

The just getting it done attitude.

We don’t just get things done and we go above and beyond for clients. If we quote £50k for an application and we expect it to take 6 months and the client is still refining 12 months later then we do what is right by the client and don’t charge more.

Again, how is the financials an argument for AI's quality?

We have local councils, one of the largest phone providers in the Uk, and other massive clients that always come back for repeat business and recommending us to other people. We do not advertise and we are not short of work.

You praise your sales people here. How is this related to code vulnerabilities or maintainability? The projects still don't get billed by those metrics.

I have to wonder how many of these people saying LLMs are useless in this context actually earn money coding and how many just rant about it because it’s the trend here.

I've started working with LLMs integrating GPT-2 (yes there was no 3 yet) into a content delivery system. The LLMs still have the same fundamental problems, but the sound more coherent, that's all.

[–] donuts@lemmy.world 9 points 6 days ago (1 children)

Except you're not using a tool, you're replacing your comment with the tool.

[–] Snowpix@lemmy.ca 3 points 6 days ago

Eat a bag of dicks. Nobody wants to read LLM garbage, regardless of whether you think that's ableism or not.