Slop is tolerated in the enterprise space because there is a business entity behind it
Posted by ChiefAoki@reddit | ExperiencedDevs | View on Reddit | 115 comments
I'm not talking about AI slop either, I work for a pretty big conglomerate and have transferred internally through numerous acquisitions throughout my career. Every single organization I have ever had the displeasure of working for, has their flagship product running on sloppy spaghetti code written by people who don't give a shit a decade ago, long before AI and agentic coding was a thing.
I started wondering why, if the underlying codebase is so poor and prone to bugs, that businesses still flock to these products, signing years long vendor agreements. It wasn't until my 4th transfer that I realized that the only thing driving sales was that there was an established business entity behind said products with an in-house legal council. These business entities see anywhere between one to five new lawsuits every year, and yet, every year, revenue and net profit goes up.
It's almost mind-boggling to me that we can continue to push untested, unreviewed code to production that will have widespread consequences, and yet we don't actually have an incentive to fix our products, because other corporations like having an entity they can bring to court when things go sideways, and even if things go sideways, a well-funded legal department will just sort it out where everyone comes out on top.
We recently had an AI mandate company-wide, and there are some people who think this is going to result in more slop, but I don't think it fucking matters, because it's like pissing in the ocean.
pydry@reddit
Have you ever wondered why the tech industry has startups and other industries seemingly dont?
Like, why a small company with limited resources in tech can on occasion clobber an enormous behemoth while in, say, oil or construction they cant?
This is why.
baezizbae@reddit
Elaborate more on this?
pydry@reddit
Big company makes a piece of shit product and then gets fat and lazy.
Startup creates way better product with a team of motivated devs who build product that steals their lunch.
It isnt the story of every startup nor does every fat and lazy corporate behemoth lose customers over their piece of shit software but it happens - way more than in most industries.
it's not as simple as "build a better product", the marketing and sales strategies need to be good too and everything has to be well executed.
thy_bucket_for_thee@reddit
This is neglected that most startups are being propped up by speculative finance but using literal pension funds.
I'm sorry but the VC model of tech innovation is beyond pathetic. The public should have a right to decide the direction of tech in a nation, not a bunch of goons that cream themselves for every worker they crush.
pydry@reddit
im not making a value judgment about whether piling money into tech startups is socially valuable. on the whole i think it isnt. nonetheless, the ability of a small team to disrupt a big business exists where in other industries it doesnt, in large part because tech debt actually does matter.
oldsecondhand@reddit
It's not just tech debt but also management debt that allows disruption. Too high headcount can make any change slow. You get longer meetings, more stakeholders and more office politics.
pydry@reddit
This is also true. Organizational dysfunction also creates startup opportunities.
BetterWhereas3245@reddit
Organizations mimic us, they suffer from cancer too. They grow with tumors, sometimes the tumors become the majority of the company.
Tumors in the software, tumors in spending, tumors in management, tumors in process.
The level of dysfunction in the corporate world is appalling, and don't even get started on the government sectors. How does anything even work? It defies all intuition that society even manages to function.
ExpWebDev@reddit
There are still startups that don't get pushed with the VC backed growth of "put as many 'bums on seats' as possible to make side products that are more useless than the actual thing we built on". Some competing products differ in head count by orders of magnitude. Snapchat with ~5000 employees and telegram with like 30. And both have MAU in the hundreds of millions.
GeneralBarnacle10@reddit
Yeah a big reason why new tech businesses are the way they are is because the potential return is so large in software. You aren't going to get 1000x ROI and $1B+ valuation in consumer package goods.
CyberDumb@reddit
Start-ups are free R&D for big corporations. They rarely clobber a big player, they mostly get bought. That way corpos spend money only for the successful projects.
Hziak@reddit
More than that, I think demand drives business in these cases. There’s a demand for someone to upset the big businesses. I’ve never met someone happy about ServiceNow in their company, or Salesfoce, and yet they’re both everywhere. People would flock to the first viable alternative that offers an easy migration path out. Other industries don’t have that in quite the same way - the power in contract negotiations usually lies with the consumer, but in software, it often ends up in the hands of the service provider. So there’s a huge potential market for disrupters right now that’s pretty unique to high tech.
The problem is that large corporations don’t want to take chances, so many of these startups are acquired or fail before they develop enough of a reputation to compete for the contracts. Hence there’s still no significant SNOW/SF competitors despite many attempts.
ComputerOwl@reddit
Viable is pretty vague but I don’t think that. People who make the buying decisions just want something that doesn’t get them in trouble. The decision to use the same software everyone else already uses is less of a risk for their personal careers than to choose the newcomer, even if the established software is garbage.
BetterWhereas3245@reddit
How did the old adage go? Nobody ever got fired for buying IBM?
We as engineers tend to believe that quality matters, that people care about their "craft". But in the world of software, only we are partaking in anything remotely resembling a craft. Everyone else is doing a 9 to 5 office job, they really do not give a single fuck about anything. The sooner people understand that, the sooner everything else starts to make sense.
Companies are motivated by short term profit, software is a cost center, quality is incidental or a nice to have last on a long list of things that take precedence in the eyes of management and ownership.
You will be paid to ship shit, and you'll be happy.
Hziak@reddit
I admit, viable is a HEAVILY loaded term here, but I think companies are sick of being hostages to annual rate increases and have a very bad taste in their mouth from these service providers. I’m not convinced that everybody is as cautious anymore. The senior leadership turnaround has decreased in time to a few years at most companies these days. So I can’t imagine that taking a risk that would result in a strong short term performance, such as cost-cutting would be something that not a lot would consider. I would say the big hurdle would actually be that the project would probably take such a long time. Which is why I emphasized the first company that has a good migration plan and convenient tools for it.
ChallengeDiaper@reddit
Exactly. Startups move fast, but companies move slow. What works gets acquired.
pydry@reddit
they often get bought for an amount of money that lets the founder retire in luxury and then killed.
if the big corporation could replicate the magic without shelling out a big pile of cash they absolutely would. it's not "free".
CyberDumb@reddit
They sometimes use the project as is and make it "production" ready.
Vegetable-Ad-7184@reddit
I think this is a very idealistic take, which comes from an admirable disposition, but, is mis-placed.
Theoretically, a focus/optimization on quality exists in all industries.
What is ~unique to software is that the marginal cost of selling the product is close to zero, which is not the case for physical goods.
Music has up and comers. So does literature, comedy, film. Essentially, any intellectual product that can be copied as a form of distribution... and quality doesn't have to be a big factor, just finding a market. Jeff Dunham is not a quality artist, yet, he has million$.
Ultimately, this is the reason for both bugs and unicorns in tech - pushing shit is free.
awkward@reddit
It's not that none exist, but startups very rarely win on quality unless they're up against a really awful incumbent.
It's more likely to go the other way - the startup is able to succeed by creating a disruptive product. They're focusing on the core problem and ignoring less useful features or concerns that were developed for niche use cases.
edgmnt_net@reddit
Companies rarely build high impact projects, many of them don't even have a product and it's more like low cost custom software development in disguise. I blame macro stuff, there's a lot of cheap money to pump around.
pydry@reddit
OP is complaining about really awful.
Im using one right now it's called figma.
Turnip-itup@reddit
I think this take discounts the uniqueness that tech has. Like assuming that competition from incumbents is same, tech allows a small team to work at a scale that say oil or manufacturing simply can’t . There’s too much upfront cost required for any new startup to start a business in oil or steel products .
pydry@reddit
i didnt discount those things. they matter too. all the stars have to align it just so happens that good tech (meaning low tech debt) is one of those stars.
SpritaniumRELOADED@reddit
In 99% of software, bugs are considered acceptable. There is no business need for refactoring, and when the codebase becomes impossible to work on, a rewrite is done to start the process over again. The whole thing is much less of an issue than people make it out to be
DigmonsDrill@reddit
For the most part, it doesn't matter if the code is garbage if it works.
It was already a blackbox to management before. It will be a blackbox to management after AI.
edgmnt_net@reddit
This is why they hire engineers. Management isn't supposed to micromanage things.
DigmonsDrill@reddit
They really don't care.
There are still a few people mad that compiled languages generate garbage assembly code, assembly they couldn't possibly understand.
I don't know for sure that's where the industry is going, but it could be. That for 95% of cases, no one cares what the source looks like.
There will still be life-critical software, like in cars or space ships, that needs to be written extremely carefully.
djnattyp@reddit
It's "acceptable" to the business dingdongs running it, but screws over the actual users of the software. Just like insurance companies - that are great until you actually need them.
SpritaniumRELOADED@reddit
The business dingdongs are the ones funding the development of the services
PrydwenParkingOnly@reddit
Yup. Many mid-level or early senior developers do not understand the business case. Just like the creator of the post, who thinks the business thrives because of lawyers…
If you spend too much time on stuff that does not drive up sales, then you’re doing the wrong thing. IT is in support of the business. IT is almost nowhere a driver on its own.
SpritaniumRELOADED@reddit
I feel like a lot of people treat their code like an arts & crafts project and that's why they're so upset that it's become easier to produce
djnattyp@reddit
Well, the actual users (sometimes through circuitous routes) are providing the funding to the business dingdongs.
SpritaniumRELOADED@reddit
Yes, and they have a tolerance for bugs.
Manic5PA@reddit
In this age of SaaS, by the time end user revenue is funding anything (if it ever does), the product is already a decade+ old to begin with and all the technical debt is already in place.
failsafe-author@reddit
Of course there is a business need for refactoring. This is like saying there isn’t a need to get an oil change in your car.
Manic5PA@reddit
The problem is that the rewrite usually never comes, and if the company isn't big enough to start making acquisitions (or makes the wrong ones) then it either goes tits up or (more likely) gets cannibalized for its client base.
josh_bripton@reddit
I’m inclined to give this position a chance because I see so many apparently smart people endorse it. But how do you square this with the fact that migrations take enormous headcount and often drag on for many quarters or even years depending on the complexity? It could be that “this can’t be avoided even if people have a shit”, maybe that’s why the status quo strategy is so widely agreed upon. But it can’t be that “rewrites aren’t costly”, they obviously are.
AnInstant@reddit
The obvious bottleneck of most systems/companies are obviously Product Owners/Managers who are not fit for their role and they can't even write good AC for their tasks. So things are developed as trial and error, lots of dead code, feature flags and other things because manager can't decide which option he wants so he wants all of them being switchable, then never giving time to clean up this mess because "we need to push this and that". It's not like rewrites can be avoided in this cycle. Rewrites become mandatory as it's easier to start fresh with knowledge we have than to refactor this crap. Obviously there are systems too big to rewrite and they will just live with layers of crap design decisions and dead code.
Many people would disagree with me, but best work I had in this industry was to detect the right time to decouple/rewrite projects/modules so I avoided scenarios when it was too late and it had to be legacy 4ever. Worst were situations when customer thought it's cheaper to keep going with legacy till they could barely find devs to touch this crap and they paid way more for devs than rewrite would cost years ago. Of course, you need the instinct when it's worth and when it's not, I had projects when it was obvious rewrite would cost more than this app is worth and ever earn.
Franks2000inchTV@reddit
Rewrites don't need to be total. Migrations can be done piecemeal.
At some point the benefit outweighs the cost.
SpritaniumRELOADED@reddit
I think the main delimiting factor is whether you're in an industry that assigns economic value to stability. Systems that support hospitals, factories, etc. have been in place for decades because they were built carefully with that goal in mind. Systems that support media streaming and todo lists can be handled very flippantly by comparison
lurco_purgo@reddit
That's what I thought, until I joined a company in the aerospace sector. The absolute wost code I've ever seen, barely functional.
Maybe we would have a bit more time to improve the code, but coroporate workshops with HR where you're asked "what quality means to you" are probably more important...
catch_dot_dot_dot@reddit
It's funny when people think aerospace, medical, etc is such high quality because of safety considerations and that's probably not the case. If you love Ada so much, why can companies not find Ada devs if their life depended on it. Hint: it sucks in the real world!
Sunstorm84@reddit
Having seen the code for some of those systems, nope, they’re also mostly human-written slop littered with bug fixes and hacks from over the years.
catch_dot_dot_dot@reddit
Yeah, it's the extensive manual QA that defines these systems. The code is often average. Not great, not terrible, and made worse by years of building on top.
ciynoobv@reddit
I’d argue that they are often worse because those industries generally see software as a pure cost-center. Additionally the amount of red tape means that iteration cycles are generally measured in months which means issues stay unfixed since they get discovered late and it isn’t viewed as important enough to warrant dealing with the whole process in order to fix them. I really fucking hate "health-tech".
josh_bripton@reddit
The counterfactual I have in mind is just “constant refactoring, enforce minimum code quality standards, never rewrite”. I get that bugs often don’t matter, but burning huge man hours on migrations seems like a dumb thing that is endemic in many areas of software engineering and I don’t have an explanation for it. I’m not saying that everyone should write hyper abstract “ready for anything” code, but noticing that things are a mess and taking some sprint time to do a refactor when you don’t have a huge deadline seems kind of easy to do. Saying “oh boy our only option is to rewrite” seems like a very clear case of poor stewardship and actually a terrible strategy for shipping features quickly.
zuilli@reddit
I think it's more of a "rewrites can be avoided if everyone involved gives a shit" but that condition never happens in companies so we have no use for working with that unreachable ideal and should start dealing with things pragmatically.
moh_kohn@reddit
Notably all these businesses rely on operating systems and libraries that are maintained very carefully over decades.
hiddenhare@reddit
Leaders who enjoy rushing out new features today would also like to rush out new features in six months, without being heavily delayed by a rewrite. Technical debt makes the people in charge powerless, because it makes everyone powerless; it’s common for large enterprises to have an engineering team who say “no” to every request!
Antoine8811@reddit
This has always been the case but feel like is being amplified by code gen AI.
Like you said, high functional tooling to service the customers need has never been easier to achieve. Hard to compete on just that.
Having a high-reputation and/or to-big-to-fail type entity behind the tool can't be easily built and becomes big a differentiator.
edgmnt_net@reddit
Yep, and projects have been kicking the bucket like crazy in the past decade. Just look at how much stuff Google killed. I suspect a lot of that stuff dies after accumulating tech debt and becoming too expensive to push forward. This is good ol' debt and leverage, probably worse. With AI you get even more such leverage, orders of magnitude even, considering vibe coding.
ng37779a@reddit
It's not just about having someone to sue — it's the switching cost nightmare. Once you're locked into their workflows, vendor integrations, supplier networks, and internal processes, ripping it out becomes a multi-year, multi-million dollar project.
There's literally an entire industry around ERP migrations. Companies paying consultants millions to move from one pile of spaghetti to another, slightly different pile of spaghetti.
But, the brutal truth is that building something that isn't slop would be incredibly expensive and take years. Which is why there are maybe 1 actual quality enterprise player in any given space, and they charge accordingly. Everyone else just races to "good enough to get the contract signed" and banks on switching costs to keep customers trapped.
The AI mandate probably makes it worse because now you can ship even faster without the pretense of quality. But like you said, pissing in the ocean.
PinEnvironmental6395@reddit
AI was never the real problem, it just exacerbates the real problem: slop. We talk about slop as an exclusively AI issue, but it's absolutely not. Commercial software shops have always been slop factories. As AI creates more and more slop faster than ever a lot of developers are learning very quickly the virtue of good old waterfall software engineering. That's what all of the coding agent harnesses that scaffold plan/build/review around an LLM really do - enforce something resembling waterfall SDLC. And they work well because methodologies where you stop and think and review aid adjust work well. We ran away from waterfall because it was slow and cumbersome bureaucracy for hackerish engineers, but businesses ran towards things like XP and Agile because they allowed the slop trough to be filled faster.
Life_Rabbit_1438@reddit
Companies don't make decisions. People do. What's best for the person making the decision often has no relevance to what's best for the company.
Low level people care about what is best for the company. Managers get ahead by limiting personal responsibility regardless of how that impacts the company.
So outsourcing to low quality vendor reduces personal responsibility.
dom_optimus_maximus@reddit
so you've worked for Cisco I see.
Foreign_Addition2844@reddit
Humans were writing slop code long before AI.
Obsidian743@reddit
Software is a tool to meet a business need. End of story.
We have a name for the perpetual employees who never seem to get it: Senior Engineer
sundevil21CS@reddit
Idk what big company you guys work for but at my F100 company we have important systems that I would consider technical debt and spaghetti code written 100% by humans.
It just takes months of getting that code to production extensive testing so even though the code itself isn’t great it still doesn’t break.
I think AI is a net positive better option than this honestly.
Varrianda@reddit
Dude, when I see people say “ai writes slop code” it drives me insane. I can count the times I’ve been in a codebase and said “wow this is elegant” on one hand! There’s no such thing as “good” code unless you work at a company that heavily invests in good engineering.
jexxie3@reddit
Agreed. It is almost like if we invested in people instead of rotating contractors in and out every 3-6 months who are pressured to push out features as quickly as possible without any underlying understanding of the codebase or business logic…
And garbage architecture for garbage reasons…
I just migrated a batch job to some piece of shit low code platform just so that we could say we are using an automation tool that we are paying god only knows how much for… that we purchased so that we could… idk? Get around change management??
Surprise! The batch job hasn’t even run in 2 years.
sundevil21CS@reddit
Garbage architecture for garbage reasons is a crazy underrated point too. The amount of times I’ve seen poor architecture choices deliberately be pushed through due to things like budgeting priority across teams, pivots to save face with leadership, or to make something an easier sell to product or marketing.
It made the whole system design space and decision making frustrating because moving to more senior roles you learn all about technical trade offs, but at big companies those decisions are more about politics while tolerating enough for the system to still function at scale than actually building the best system.
fued@reddit
most architecture choices are because someone wants a tech on their resume.
bigtimehater1969@reddit
There absolutely is such a thing as "good" code in a relatively objective sense. Abstraction layers are objectively defined (in code), and can be judged. Not all algorithms are equal.
The main issue is a large majority of engineers aren't good enough to discern "good" code versus bad code. Most engineers look at code, realize they can't read it, then call it bad code.
Rating code requires you to firstly be good at reading code (a lot of engineers who rate themselves "good" are already disqualified), then understanding the context and requirements under which it was written. If code was written for different requirements than yours, then no shit it's not going to do what you need it to.
And no, every line not having a comment does not mean it's "bad code." Not using your favorite design pattern, language, or naming convention does not make it "bad code" either.
Antique_Pin5266@reddit
Human bad code when not written by outsourced resources usually is just a record of business decisions made up to that point
Whereas for AI slop there is no rhyme or reason
Ibuprofen-Headgear@reddit
The main difference for me is I could usually tell the intent from “bad code” because it came with a certain style or pattern of “badness” or issues, whereas the generated slop is slop in a different way that’s harder to derive the original intent/spec from. Not to say that a combination of ai plus human can enhance that or alleviate that, it’s just the issue I have with nearly-unreviewed ai code being shoved everywhere.
Inb4 people problem, company culture problem, etc; yes, I know, literally every problem is those 2 things basically
rsalot@reddit
If there is no such thing as good code, then there is no such thing as bad code
No way that any code is equal
Varrianda@reddit
Thats not what I meant. Most code is just…bad code, or at the very least “passable” code. AKA just as good if not worse than agent output.
syklemil@reddit
Yeah, we've been complaining about shit code for as long as there's been programming, but I think a lot of us also see that kind of AI bro as this old Invader Zim clip.
Or possibly someone bringing an excavator to the law of holes. Or the old four types of officers category "stupid and industrious". There are so many adages about this kind of situation.
In any case, the people who can barely hold a programming language the right way up aren't going to produce better code with an LLM—they don't know how to recognise it in the first place. They're going to deskill, and produce way more content for thedailywtf, way faster.
anonyuser415@reddit
"most code is fine and AI writes code that is fine" just doesn't have the same ring to it
raralala1@reddit
The question is can the AI continue refining the slop, behind that spaghetti code someone in the back probably understand that code and can fix it,
Also I think people doesn't understand what slop is, spaghetti code isn't slop, slop is quick, easy and overabundance to made product/code.
seanrowens@reddit
Gotta be ffmpeg
AnInstant@reddit
People don't get that you can control what AI produces and just be faster developing things, before you could spend months reading spaghetti and still not being able to work with that. I once had a project decompilled from binaries that had variables like a, b, c etc. without any explanation - it also happens when someone is so stupid to delete repository with code. And it was actively maintained by few devs as a critical application, every change took weeks of testing what this code actually does. AI will sort things out in minutes.
Dense_Gate_5193@reddit
most of the time you can have it just optimize the code in place assuming you have unit tests hardening the behavior. then you just have it directly 1-1 copy it to your new favorite language.
fued@reddit
simple fact is, slop is better than half the garbage my co-workers create.
Tcamis01@reddit
This mirrors what I've generally seen also. The human slop is generally worse than the AI slop.
Polus43@reddit
I don’t have evidence, but I think you’re greatly underestimating nepotism/“networks”/bribery/corruption.
One of the most advantageous parts of dealing with senior management as large multinationals is they can create jobs for almost anyone.
Have seen deal calls numerous times talking about “hiring X”
FastHotEmu@reddit
When the user and the customer are different people (which is very common in Enterprise) the user suffers
ExpWebDev@reddit
Companies can be looked at like fortresses. Lawyers are part of front line defense and PR and marketing are front-middle. Product at the center. But there's not much incentive to make the product very strong if you're confident your outer defenses are strong, as your 2nd to last paragraph tells. If they're doing a good job curtailing liability buggy software then nothing else matters to them. Also it's easier to talk up building legal and PR strength, than improving software integrity with non technical people
goship-tech@reddit
The cynicism about org incentives is warranted. But I'd push back on the idea that individual measurement is pointless even when the org won't act.
Knowing whether your own AI habits are good or quietly degrading is useful regardless of what management does. The devs I've watched stay sharp while using AI heavily all have high accept/reject discipline — they're opinionated about when the model is wrong. The ones who drift tend toward high acceptance rates and batch testing as shortcuts. That pattern shows up before output quality drops.
Hard to see in your own work without something external pointing at it, which is part of why I built goship.tech.
Aleks_Zemz_1111@reddit
This hits hard. I run precision industrial machinery for my day job, but spent years writing React before realizing the tech industry is exactly the same as the factory floor. Management doesn’t care about the 'craftsmanship' of the code; they care about the output quota and who is liable when it breaks.
You nailed it: they aren't buying software, they are buying an insurance policy with an in-house legal team attached.
It’s wild how many developers burn out trying to write 'perfect' code for enterprise systems, not realizing they are essentially just acting as highly-paid janitors for someone else's legacy code. An AI mandate won’t break the enterprise because the system is already built to tolerate polite noise and AI sludge. Code is a commodity. Ownership is the only thing that actually matters.
coderstephen@reddit
This hurts me.
For me to get enjoyment out of software engineering, I have to make that so-called "perfect code". It's what's enjoyable, I like the challenge.
The change has been that 10 years ago, people appreciated such code even if it wasn't necessary. And eventually I got tasked to work on the most critical modules of code where correctness mattered, because they knew I'd take great care on it.
Nowadays with prevalent AI code generation, it seems like people don't appreciate code any more. It's just a boring stinky chore, what do you mean there's a difference between good and bad code? Who can even tell? It's like an architect who can't tell the difference between a beautiful building and a concrete rectangle because technically, they are both buildings and get the job done of holding people.
max123246@reddit
People still care about code. Higher ups just get to believe the lie even more now that speed is all that matters
bigtimehater1969@reddit
It's not just the legal aspects. You have to consider how much does it cost versus how much value does it bring to "do it right."
The honest fact is, so many software engineers are completely divorced from the business aspect of software that they don't realize doing it "right" all the time usually ends up bringing less value than it costs.
Ship fast but ship a lot of bugs? Guess what, unless you have an SLA the honest truth is your customer will grumble but in the end forget about it, and your management won't really care. Meanwhile, if you ship slow with less bugs other businesses who ship fast will eat your lunch.
And quite frankly a lot of software engineers don't actually know what it means to do it "right." Most software engineers think doing it "right" means giving them an unlimited deadline to over engineer whatever they want, and hyper fixate on metrics like code coverage or something. That always ends up bringing less value than doing it fast.
ub3rh4x0rz@reddit
The more cynical view is that the more capital, the more regulation, the more purchasing bureaucracy among target customers, the less churn and competition. You don't need to deliver UX that sparks joy in these contexts, it's mostly KTLO.
wmichben@reddit
Because so many people live by "Move fast and break things", my motto is "All software is shit". I think people stopped caring about making software that is actually useful and consistently functional years ago. Some of the most popular software out there has had the same bugs for years and I am amazed they never get fixed. The business of software can be pretty sickening at times and I'd love for there to be a sea change where people start caring about quality and serving the user again.
zamkiam@reddit
alot of slop doesnt effect business process or revenue in the mind of the slop creators
mmcnl@reddit
Large business are organic organizations that need bureaucratic guardrails to function. One of those guardrails is to mitigate risks with extensive procurement processes. So they only deal with vendors who are willing to go down this cumbersome path. The quality of the product is actually not that important as long as it ticks requirement boxes.
fried_green_baloney@reddit
I think one of the reasons the LLMs are supposed to replace programmers is that the standards of quality for most programming tasks are abysmally low.
I was a close bystander to a company selecting a database product - this was before open source RDBMS systems were a thing. The arm twisting, free meals, free classes, etc etc, was jaw dropping to see.
On the other hand I did get a couple of nice free lunches out it.
In the end they ended buying a not-Oracle selection for about $50,000 to start. Today of course you could spin up PostgreSQL or MySQL/Mariadb in an afternoon for free.
invest2018@reddit
What an appalling equivocation.
coderstephen@reddit
Everyone's software is buggy, so your customers expect and tolerate buggy software. So business reasoning is why rise above your competitors by a noticeable degree, when instead you can be perfectly profitable by acquiring customers through other means. Better price, better marketing, better partnerships, etc. Then they pay for your buggy software that is no better or worse than anyone else's.
Professional-Dog1562@reddit
What happens when your system performance suddenly drops because users are scaling quicker than refactors can happen? It's happening right now for my company. Product won't let me come off the gas pedal and our systems are degrading. It's going to be at a standstill soon.
Zetus@reddit
Complexity maintenance is the whole name of the game in software, teams that have poor decisions will compound and limit their own tech.
Potential4752@reddit
That doesn’t make sense mathematically. When two parties go to court, one or both of them is going to lose money.
Skullclownlol@reddit
This view is much too narrow. E.g. airline Qantas was fine $90M after being sued in court: https://www.abc.net.au/news/2025-08-18/qantas-fined-in-federal-court-job-outsourcing-penalty/105659978
Except they made an estimated $500M profit off of outsourcing illegally, so even after getting fined $90M they still made an estimated $410M profit.
So both the fined and the courts made money. It's the third party, the illegally replaced workers, who paid the price.
Potential4752@reddit
In your example the two parties involved are not doing business with each other. The damaged party was not in court. If they were, someone would have lost.
Skullclownlol@reddit
Potential4752@reddit
You’ve got to use context. I also didn’t specify that I meant a US court. Are you going to accuse me of moving the goalposts if say that Albanian courts don’t count?
The actual damaged party is the workers who were not hired, not the ACCC. The workers did not win when compared to the null state of Qantas following the law.
ChiefAoki@reddit (OP)
Corporate A takes Corporate B to court, Corporate B offers a discounted rate to Corporate A for a year or two in return for the lawsuit dropped or settled out of court. The discount is just enough for Corporate A to justify continue using Corporate B as a vendor, and maybe even act as an affiliate in the future. Few years down the road, no more discounts, same old shitty service, but Corporate A is so entrenched with Corporate B's services that they can no longer go anywhere else, so they keep coughing up. Corporate B keeps making money from Corporate A and have a lifelong customer.
When settled, Corporate A is usually also required to sign a non-arbitration requirement which means they can no longer take Corporate B to court in the future.
CaffeinatedT@reddit
You've just stumbled into one of the many reasons why developers who are dismissive of every non-technical consideration are nearly always wrong about everything long term.
massive_succ@reddit
Absolutely this. Developers focus too much on the technical issues they can see, and not the business and regulatory ones that they aren't exposed to. We as developers have to start thinking about the boundaries between companies serving as legal/regulatory entities rather than strictly technical ones.
To use a non tech example: If Uber wants to offer drivers cars to rent to use for Uber, naively you might assume they'd buy cars and then rent them to drivers. But no, they partner with a company that just rents cars, and have that company offer the cars for rent, facilitated by Uber. There's no technical barrier to Uber buying the cars and renting them out themselves. But they do not want the liability associated with upkeeping the car, and they want the ability to fire and replace the rental car company in the future. There are also accounting benefits to keeping assets like rental cars off of your company's books.
In software it's very similar. Could most companies build a simple enough set of software to run their business? Absolutely, especially these days with vibe-coding. But externalizing that responsibility allows you to switch software vendors, or sue your vendor if something goes wrong rather than take responsibility yourself. (You can play this out mentally with consumer lawsuits as well.)
In this lens it's easy to see why "slop" is tolerated: software working as designed is not necessarily the primary concern of the buyer or seller. To go back to the rental car analogy: Neither Uber nor the rental car company really cares if the cars are ugly or broken, as long as they can do the bare minimum, and the users suffer.
To make quality a central concern you either have to regulate it directly (as we do with banking/healthcare software) or you have to have a legal regime that disallows the externalization of risk.
michelle7affy2025@reddit
do you think this legal safety net will change with ai integration?
CaffeinatedT@reddit
Not unless government takes an interest in it (not impossible but in the anglosphere unlikely imo). Companies will continue to scream at each other and continue to get legal to fix it. Individual consumers will get rolled over by more and more "computer says no" systems.
Outrageous-Sort-3999@reddit
reminds me of the time when we had a whole team just for bug triage, like band-aids everywhere
metaphorm@reddit
the only code that you can be certain has no bugs is no code at all. the reason the legal entity status matters is because that's how accountability works in this space. customer doesn't care about your bug. customer cares about their recourse if they experience financial loss due to your system malfunctioning.
Extra-Organization-6@reddit
switching cost is the real answer. once an enterprise has built 15 years of business processes, compliance audits, and team knowledge around a crappy codebase, rewriting it means rebuilding all that context too. the code is the cheap part. the organizational tax of replacing it is what keeps the spaghetti running. ai just makes the slop accumulate faster, doesn't change the underlying economics.
thekwoka@reddit
As much as the "code" is the product, the REAL product is marketing and sales.
That's like 99% of what you're actually paying for with all that software.
WJMazepas@reddit
Business still use these products because the bugs arent annoying or bad in the eyes of the executives whom decided to contract.
So many times, a company starts using a software, everyone who effectively uses it, hate it but the executives don't really care about that. They simply heard that the software is good, so surely the issue has to be in the users
MonochromeDinosaur@reddit
Even if your business is an internet business the only thing that matters is cold hard sales.
That’s why garbage legacy apps still make tons of money. Too big to fail, too entrenched to change, too spaghetti to improve.
I’ve learned to live with it. Just put me on the highest visibility highest revenue capture project and pay me handsomely.
The slop will happen regardless of effort trying to avoid slop is a sisyphean task.
WJMazepas@reddit
Your comment appears 3 times in this post
MonochromeDinosaur@reddit
Reddit 🤦🏻♂️
MonochromeDinosaur@reddit
Even if your business is an internet business the only thing that matters is cold hard sales.
That’s why garbage legacy apps still make tons of money. Too big to fail, too entrenched to change, too spaghetti to improve.
I’ve learned to live with it. Just put me on the highest visibility highest revenue capture project and pay me handsomely.
The slop will happen regardless of effort trying to avoid slop is a sisyphean task.
MonochromeDinosaur@reddit
Even if your business is an internet business the only thing that matters is cold hard sales.
That’s why garbage legacy apps still make tons of money. Too big to fail, too entrenched to change, too spaghetti to improve.
I’ve learned to live with it. Just put me on the highest visibility highest revenue capture project and pay me handsomely.
The slop will happen regardless of effort trying to avoid slop is a sisyphean task.
dude1435@reddit
code make money or code dont make money
Groove-Theory@reddit
shitty software make line go up