I wonder if the flood of vibecoding startups may end up helping long established enterprises and OSS through eroding consumer trust towards new software, and highlighting the value of the industrial software development process. Bills of materials, 4-eyes code reviews, release gates, QA testers, planning involving domain experts, external audits etc. Any time I see a new large piece of software nowadays that seems to have popped out of nowhere and has generic sounding docs and testimonials with em dashes I tend to wonder if it was generated with a one shot prompt and may include vulnerabilities, malicious code, or business domain hallucinations.
It's a bit like with books: 10+ years ago I could have trusted a book from a new author I never heard about just because the large effort of getting published was hinting that the author is likely qualified. Now, hell no.
When the barrier to publish drops to zero, the barrier to gain trust skyrockets.
How are you so sure that big corporations will also keep doing code reviews, QA testing, planning. At the current moment, they are laying off a large portion of their workforce, forcing devs to take more and more ownership and even expecting devs to assume the QA testing role. I dont think any software is trustable anymore.
Fair point: perhaps the true advantage of long established enterprises and OSS orgs is not the process, which can be considered private implementation detail, but the public track record. Either way, startups don't have that.
It's all about responsibility. If I pay a company for functional software and it doesn't function, that's on them.
I expect a much greater emphasis on contractual performance and liability now that the customer has even less reason to trust that a vendor is doing the right thing.
The real question is whether these security practices can truly survive the onslaught of 24/7 attack swarms, or if we're going to have to rebuild the entire cloud from the ground up.
The essence of business is not that anyone who can write code is guaranteed to make money. In the past, software companies made money because they possessed the ability to help enterprises complete digital/automation and solve problems. Writing code is just one way to solve problems. The vast majority of software I see today only solves the problems of its creators; they don't care or pay attention to how others' problems are solved.
Of course, you might say: "Nowadays, LLMs are so cheap that everyone can write software." But the fact is, while everyone can write software, not everyone can write a complete software. And the assumption that "everyone can write software" is currently also incorrect. Not everyone can write software.
It does however lower the barrier of entry and increases what a smaller team of developers can achieve with the same amount of resources. There's still been a major paradigm shift in our industry.
So it probably (I mean honestly I believe anything can happen at this point) doesn't mean every business will just start writing all of their own software (cheaper doesn't mean free), the bar for building your own stuff has lowered so there will for sure be more software written for internal use instead of using some off the shelf program.
And there will be more competition among software companies because everyone can now build much more with much less resources.
When you widen the potential suppliers, and make it hyper competitive. You don't normally get lots of small competing companies. You get get 1 or 2 companies dominating the space.
Everyone just picks the best, because they can. The best often have hidden advantages like tons of funding, so it's actually subsidised.
When an enterprise buys SAP or Jira, they are trying to standardise the way they work. The software is just a crystallisation of the buisness process they are trying to adopt.
A bunch of Consultants or agile trainers will come and change how thier teams work.
The goal is to work in a better way but critically an industry standard way so your employees are more fungible.
The product is valuable because everyone else is using it, a moat like social network network effect…
I do still think there is mileage in this post AI, particularly as it changes howbuisness works, people will want to buy that change packaged. But maybe the frontier labs and thier forming walled gardens are the natural home of that package.
Choosing the right problem to solve was always very important. The code wasn’t always the hard part, but arguably it really was the hard part more often earlier on.
For example, a personal computer operating system that runs all your existing programs at full speed but doesn’t crash all the time and the ability to run a virtual machine on an x86 PC both took as long as they did to get to market primarily due to the technical difficulty, not because it was hard to find people ready to pay for them. Those kinds of hard coding problems are less plentiful, or at least less obvious today.
It definitely was. It was at least 50% of the hard part.
I remember there was a debate in my company on execution vs idea. Execution being writing code, the process that comes with it, and the idea being the understanding, market fit. Both sides had good arguments.
Hiring a software team to build an idea was extremely expensive and risky before ChatGPT. That's why most businesses in the world did not have a software engineering team.
Agreed. Cheaper code just means more competitors reach the same starting line faster, so distribution and actually knowing your customer matter more, not less. The moat was never the repo.
The tech will never be there with LLMs. They can't learn, they hallucinate, lie, cheat, steal, destroy things, break the law, etc. You can't train them out of these behaviors, it's how LLMs work.
An LLM is just a really fancy nail gun to replace hammers. Much faster nailing, but you need to hold it right to not shoot yourself in the foot.
Yes for the people who never coded, that was not the hardest part. We always had more ideamen than programmers, they could never execute anything. If code was never the hard part, vibecoding wouldn't be so popular.
I see this phrase thrown around alot and it lacks any substantive evidence. If coding was the easy part why were junior level SWEs getting paid six-figure salaries?
People who say coding wasn't the hard part have never had to refactor a legacy system without LLM assistance and it shows.
I think the pay in certain markets was mostly because companies had more money than sense. And then that spread... In most markets pay is not really that special compared to any other field with some education needed. And neither really is the work.
Its more of an opportunity at least for the next few years for startups to be a IT service provider and build as many software as per real business requirement using AI. As startups can be more budget friendly and quick to build and deliver. The earlier need to have a large team is no more a constraint for a Tech startup. But yes one needs to use a harness or ways to make sure they can build software using some best ones in the market.. I use Atlas.
I think it’s completely inverted actually. The big companies have a much smaller moat now. This is almost certainly the best time to do a startup. You can build a decent product very fast, GTM in a month or so, and pivot rapidly. But it requires a lot of skill to do it with AI tooling in a way that sustainable and the techniques and tools are very immature and shifting around rapidly. Large companies are struggling very hard to integrate these into their flows and workforce while a small company, with the right small set of people, can achieve as much or more than a large one with thousands of developers.
I had the same thought - that its end of software startups since most people can develop software on their own now using LLMs. But seeing how people use LLMs to develop the app made me realize that there is a difference in knowing how software is built vs just vibe coding it.
For example - only a good s/w engineer would know to include CI/CD pipeline which will help her deploy changes faster. A person who is not familiar with s/w is not gonna do that unless they dig a little more deeply OR follows an advise from a good model. In 2 months time, the non s/w engineer is gonna drown in keeping up with all the pieces of the product without regression.
So startups who provide value in short cycles would still be in demand. The cycle time of requirement -> delivery is gonna be very short.
A premise here that doesn't stick, "users are expecting open source and free software" this is definitely dependent on the type of software, and reddit isn't even the best way to demonstrate this reddit frowns at everything anyways. There are definitely still a lot of new startups making revenue it's just harder these days.
One a much larger time scale, I think that premise still holds.
Greybeards used to tell me how back in the 80s/90s "everyone" (in software) was developing & selling their custom solutions. And it was completely normal to buy your software, the OS, IDE, compilers, Browser, basically everything. But with the proliferation of free software/open source many software business models had to compete with viable free alternatives.
And with big tech doing price-dumping in many markets (who still buys a browser?).
So nowadays many software sub segments there exist very good solutions that are free, competing with those are super hard and only possible for very specialized niches.
If you are settled with distribution channels, can productize the solution or convert the startup feom product one to a service model then you are probably safe.
The development itself is cheap and has no moat.
If you solve a problem, a real problem, and has the right niche for it, for sure it can survive.
AI just speed up the software part that will also rise the amount of features that was impossible in the past without a huge numbers of employees.
To me, it seems that the modern software startup is geared up to provide highly customized solutions to customers. Software that easily bends to the customers needs, rather than customer changing their processes to make use of software.
The environment of computation and it's interactions in the 7 day week of a modern classed society is open to debate and great thought will come from it to solve the problems of mere boredom , as well as the vital life sustaining things people need.
I find myself building far more simpler single use tools that do exactly what I need than using commercial swiss army knives which usually overcomplicate what I need to do.
This is a qwuestion that has been on top of my mind for quite some time now.
My answer is it depends so munch on what are you trying to sell and to whom.
As software engineers we see a potential to vibe code any product that we see, but we are alone in this, imagine the guys working as BAs or other roles at multinational corporations. Most of them have absolutely no ideea how to even vibecode a working app, that is only easy to us as engineers. So you can still atract new customers, it's just about who you target and that is mostly it.
True that your products will get cloned, but so what? If people start cloning you, that is only a proof that you are on the right track. Keep on cooking guys
Creating new software is far more than just a matter of tokens. LLMs are like a small programming team: you still need a project manager (you!).
And tax return software is a perfect example of where "vibe coding" will go horribly wrong: a superficial tax return is trivial; handling tax rules for every different situation is vastly more complex. (1) You need to codify all the tax rules; (2) You need to do it reliably (which means really understanding how to work with LLMs: harnesses, gates, verification, reviews, etc.); (3) You need a corpus of real-world examples (taken from where?); (4) You need user feedback.
This would be an intensive multi-month project. And that's before even looking at marketing and sales.
Can a competitor race with you? Sure, but only if they're also determined to put in the work.
I wonder if the flood of vibecoding startups may end up helping long established enterprises and OSS through eroding consumer trust towards new software, and highlighting the value of the industrial software development process. Bills of materials, 4-eyes code reviews, release gates, QA testers, planning involving domain experts, external audits etc. Any time I see a new large piece of software nowadays that seems to have popped out of nowhere and has generic sounding docs and testimonials with em dashes I tend to wonder if it was generated with a one shot prompt and may include vulnerabilities, malicious code, or business domain hallucinations.
It's a bit like with books: 10+ years ago I could have trusted a book from a new author I never heard about just because the large effort of getting published was hinting that the author is likely qualified. Now, hell no.
When the barrier to publish drops to zero, the barrier to gain trust skyrockets.
How are you so sure that big corporations will also keep doing code reviews, QA testing, planning. At the current moment, they are laying off a large portion of their workforce, forcing devs to take more and more ownership and even expecting devs to assume the QA testing role. I dont think any software is trustable anymore.
Fair point: perhaps the true advantage of long established enterprises and OSS orgs is not the process, which can be considered private implementation detail, but the public track record. Either way, startups don't have that.
It's all about responsibility. If I pay a company for functional software and it doesn't function, that's on them.
I expect a much greater emphasis on contractual performance and liability now that the customer has even less reason to trust that a vendor is doing the right thing.
The real question is whether these security practices can truly survive the onslaught of 24/7 attack swarms, or if we're going to have to rebuild the entire cloud from the ground up.
https://www.tomshardware.com/tech-industry/artificial-intell...
The essence of business is not that anyone who can write code is guaranteed to make money. In the past, software companies made money because they possessed the ability to help enterprises complete digital/automation and solve problems. Writing code is just one way to solve problems. The vast majority of software I see today only solves the problems of its creators; they don't care or pay attention to how others' problems are solved.
Of course, you might say: "Nowadays, LLMs are so cheap that everyone can write software." But the fact is, while everyone can write software, not everyone can write a complete software. And the assumption that "everyone can write software" is currently also incorrect. Not everyone can write software.
It does however lower the barrier of entry and increases what a smaller team of developers can achieve with the same amount of resources. There's still been a major paradigm shift in our industry.
So it probably (I mean honestly I believe anything can happen at this point) doesn't mean every business will just start writing all of their own software (cheaper doesn't mean free), the bar for building your own stuff has lowered so there will for sure be more software written for internal use instead of using some off the shelf program.
And there will be more competition among software companies because everyone can now build much more with much less resources.
When you widen the potential suppliers, and make it hyper competitive. You don't normally get lots of small competing companies. You get get 1 or 2 companies dominating the space.
Everyone just picks the best, because they can. The best often have hidden advantages like tons of funding, so it's actually subsidised.
When an enterprise buys SAP or Jira, they are trying to standardise the way they work. The software is just a crystallisation of the buisness process they are trying to adopt.
A bunch of Consultants or agile trainers will come and change how thier teams work.
The goal is to work in a better way but critically an industry standard way so your employees are more fungible.
The product is valuable because everyone else is using it, a moat like social network network effect…
I do still think there is mileage in this post AI, particularly as it changes howbuisness works, people will want to buy that change packaged. But maybe the frontier labs and thier forming walled gardens are the natural home of that package.
The code was never the hard part. Understanding the market, developing product market fit, and evolving is the hard part.
Large companies aren't going to chase every opportunity.
Choosing the right problem to solve was always very important. The code wasn’t always the hard part, but arguably it really was the hard part more often earlier on.
For example, a personal computer operating system that runs all your existing programs at full speed but doesn’t crash all the time and the ability to run a virtual machine on an x86 PC both took as long as they did to get to market primarily due to the technical difficulty, not because it was hard to find people ready to pay for them. Those kinds of hard coding problems are less plentiful, or at least less obvious today.
I remember there was a debate in my company on execution vs idea. Execution being writing code, the process that comes with it, and the idea being the understanding, market fit. Both sides had good arguments.
Hiring a software team to build an idea was extremely expensive and risky before ChatGPT. That's why most businesses in the world did not have a software engineering team.
Agreed. Cheaper code just means more competitors reach the same starting line faster, so distribution and actually knowing your customer matter more, not less. The moat was never the repo.
Large companies are working hard on laying off and buying out real persons for GPU compute budget.
And then regretting it shortly after.
The tech is not there yet. It may not be ready for many years.
Keep dreaming, No CEO is regretting his multimillion dollar bonus.
They don't need the AI excuse, they did it just fine before.
The tech will never be there with LLMs. They can't learn, they hallucinate, lie, cheat, steal, destroy things, break the law, etc. You can't train them out of these behaviors, it's how LLMs work.
An LLM is just a really fancy nail gun to replace hammers. Much faster nailing, but you need to hold it right to not shoot yourself in the foot.
> The code was never the hard part.
Yes for the people who never coded, that was not the hardest part. We always had more ideamen than programmers, they could never execute anything. If code was never the hard part, vibecoding wouldn't be so popular.
I see this phrase thrown around alot and it lacks any substantive evidence. If coding was the easy part why were junior level SWEs getting paid six-figure salaries?
People who say coding wasn't the hard part have never had to refactor a legacy system without LLM assistance and it shows.
I think the pay in certain markets was mostly because companies had more money than sense. And then that spread... In most markets pay is not really that special compared to any other field with some education needed. And neither really is the work.
Yeah it was ... it just wasn't the ONLY hard part.
Distribution is the moat now :)
Its more of an opportunity at least for the next few years for startups to be a IT service provider and build as many software as per real business requirement using AI. As startups can be more budget friendly and quick to build and deliver. The earlier need to have a large team is no more a constraint for a Tech startup. But yes one needs to use a harness or ways to make sure they can build software using some best ones in the market.. I use Atlas.
I think it’s completely inverted actually. The big companies have a much smaller moat now. This is almost certainly the best time to do a startup. You can build a decent product very fast, GTM in a month or so, and pivot rapidly. But it requires a lot of skill to do it with AI tooling in a way that sustainable and the techniques and tools are very immature and shifting around rapidly. Large companies are struggling very hard to integrate these into their flows and workforce while a small company, with the right small set of people, can achieve as much or more than a large one with thousands of developers.
I had the same thought - that its end of software startups since most people can develop software on their own now using LLMs. But seeing how people use LLMs to develop the app made me realize that there is a difference in knowing how software is built vs just vibe coding it. For example - only a good s/w engineer would know to include CI/CD pipeline which will help her deploy changes faster. A person who is not familiar with s/w is not gonna do that unless they dig a little more deeply OR follows an advise from a good model. In 2 months time, the non s/w engineer is gonna drown in keeping up with all the pieces of the product without regression. So startups who provide value in short cycles would still be in demand. The cycle time of requirement -> delivery is gonna be very short.
A premise here that doesn't stick, "users are expecting open source and free software" this is definitely dependent on the type of software, and reddit isn't even the best way to demonstrate this reddit frowns at everything anyways. There are definitely still a lot of new startups making revenue it's just harder these days.
One a much larger time scale, I think that premise still holds. Greybeards used to tell me how back in the 80s/90s "everyone" (in software) was developing & selling their custom solutions. And it was completely normal to buy your software, the OS, IDE, compilers, Browser, basically everything. But with the proliferation of free software/open source many software business models had to compete with viable free alternatives. And with big tech doing price-dumping in many markets (who still buys a browser?). So nowadays many software sub segments there exist very good solutions that are free, competing with those are super hard and only possible for very specialized niches.
If you are settled with distribution channels, can productize the solution or convert the startup feom product one to a service model then you are probably safe. The development itself is cheap and has no moat.
Settleing channels are hard for small businesses.
If you solve a problem, a real problem, and has the right niche for it, for sure it can survive. AI just speed up the software part that will also rise the amount of features that was impossible in the past without a huge numbers of employees.
To me, it seems that the modern software startup is geared up to provide highly customized solutions to customers. Software that easily bends to the customers needs, rather than customer changing their processes to make use of software.
Customers are less willing to pay for software. They might try AI first on their own.
Just because I can bake bread at home doesn't mean I do.
It takes time, it takes me away from my actual job within my business, and frankly, the bread from the bakery is pretty good.
Customers will pay for something that's of benefit to them. That hasn't changed.
Make sense. Positive thinking
The environment of computation and it's interactions in the 7 day week of a modern classed society is open to debate and great thought will come from it to solve the problems of mere boredom , as well as the vital life sustaining things people need.
Some products are harder to build than others imo, even if making software has gotten easier.
Making consistent and reliable software is even harder.
I find myself building far more simpler single use tools that do exactly what I need than using commercial swiss army knives which usually overcomplicate what I need to do.
I think it comes down to a change in how companies promote and charge for the service.
This is a qwuestion that has been on top of my mind for quite some time now. My answer is it depends so munch on what are you trying to sell and to whom. As software engineers we see a potential to vibe code any product that we see, but we are alone in this, imagine the guys working as BAs or other roles at multinational corporations. Most of them have absolutely no ideea how to even vibecode a working app, that is only easy to us as engineers. So you can still atract new customers, it's just about who you target and that is mostly it.
True that your products will get cloned, but so what? If people start cloning you, that is only a proof that you are on the right track. Keep on cooking guys
Creating new software is far more than just a matter of tokens. LLMs are like a small programming team: you still need a project manager (you!).
And tax return software is a perfect example of where "vibe coding" will go horribly wrong: a superficial tax return is trivial; handling tax rules for every different situation is vastly more complex. (1) You need to codify all the tax rules; (2) You need to do it reliably (which means really understanding how to work with LLMs: harnesses, gates, verification, reviews, etc.); (3) You need a corpus of real-world examples (taken from where?); (4) You need user feedback.
This would be an intensive multi-month project. And that's before even looking at marketing and sales.
Can a competitor race with you? Sure, but only if they're also determined to put in the work.
It's still usually cheaper to pay for software than it's to write it, maintain it and run it.
- getting it promoted like nord vpn is the hard part, you do that, someone or the other ll buy something as long as it serves a utility
like bottle water