AI vs. software companies

[NnnnotSoSmart]

34:08 - Echos from the Past (Are software stocks in trouble?)

 

[NnnnotSoSmart]

Software stocks value traps?

 

AI Disruption: Moats and Value Traps
Kai Wu

May 2026

 

In AI Adopters: Beneficiaries of the Boom (Jan 2026), we showed that corporate AI adoption is accelerating as firms begin realizing tangible returns on their AI investments. As early adopters ourselves, we remain bullish on this trend and the positioning of these AI-enabled firms.

 

But disruption creates both winners and losers. The flipside of AI adoption is that many incumbents are now selling off due to fears that AI will disrupt their business models. The AI reaper has come for wealth advisors, consultants, and even trucking firms. However, nowhere has the damage been as severe and widespread as in software stocks.

 

Software development is by far the most advanced AI use case today, powered by coding agents like Claude Code and Codex. In a world where software costs nearly zero to build, the market seems to believe that software firms no longer have moats. Over the past several months, software stocks have sold off, despite the broader market continuing to rally.

 

https://www.sparklinecapital.com/post/ai-disruption?utm_medium=email&_hsenc=p2ANqtz-_C0cyv7AnoAxhy0d1VSfvHZ392thJ1XmZvM37u6rxM3CX-R9NQLUT38zLsdoNBqLbA81i2cfiGe3uhLq-7ZGzmuYCe6Q&_hsmi=419705527&utm_content=419705527&utm_source=hs_email

 

Edited May 22 by NnnnotSoSmart

[Spekulatius]

Can someone point out a single software business (SaaS or VMS) that has been disrupted by vibecoding or AI agents so far? I follow a bunch of software co but I don’t know any where this is clear.

We are entering year 3 of the AI revolution so you expect to see something.

Edited May 23 by Spekulatius

[Libs]

41 minutes ago, Spekulatius said:

Can someone point out a single software business (SaaS or VMS) that has been disrupted by vibecoding or AI agents so far? I follow a bunch of software co but I don’t know any where this is clear.

We are entering year 3 of the AI revolution so you expect to see something.

 

Does Chegg (CHGG) count?

[Spekulatius]

On 5/23/2026 at 12:03 PM, Libs said:

 

Does Chegg (CHGG) count?

Not exactly, because it is not a software business . Quite frankly, I think Chegg would have been disrupted no matter what even if AI had not happened.

Edited May 24 by Spekulatius

[Buckeye]

Interesting article in this weekend’s WSJ from Christopher Mims titled, “The AI Superstars Who Say a ‘Vibe Slop’ Crisis is Coming.”  I’m surprised it hasn’t gotten more attention with regards to the Saaspocalypse.  

 

The article starts….”Two engineers who built the core of the massively popular OpenClaw AI agent have a stark warning: The artificial intelligence supposedly capable of replacing well-paid software developers is flooding the world with bad, potentially even dangerous, code.” 
 

“You have infrastructure that’s falling apart, and you have software that’s now very, very buggy compared to before,” says Mario Zechner, creator of Pi, the agentic harness inside OpenClaw. “We can play this game for a couple more months, or maybe even years, but eventually it will catch up to us.”

 

The article concludes with…”Zechner believes a reckoning is coming. He thinks big companies will soon realize that their overemphasis on AI-produced code is driving up costs and leading to subpar software. ”

 

https://www.wsj.com/tech/ai/vibe-coding-slop-ai-tools-e6a99394?mod=author_content_page_1_pos_1

Edited May 26 by Buckeye

[nsx5200]

4 hours ago, Buckeye said:

Interesting article in this weekend’s WSJ from Christopher Mims titled, “The AI Superstars Who Say a ‘Vibe Slop’ Crisis is Coming.”  I’m surprised it hasn’t gotten more attention with regards to the Saaspocalypse.  

[...]

https://www.wsj.com/tech/ai/vibe-coding-slop-ai-tools-e6a99394?mod=author_content_page_1_pos_1

I like this part of the article:

"Catherine Wu, head of product for Claude Code at Anthropic, says the visual flicker was a side effect of a software team moving at a rapid pace, [...]. The flicker has largely been fixed, she adds."

 

The implication is that it is still not fixed, nor have they determined the root-caused of it to fix it.  Probably because they asked AI to fix it, and AI mostly fixed it.  Not saying it won't improve, but that's probably a good summary of the current state-of-the-art AI vibe coding.

 

Seeing that AI is starting to break the open source model. I suspect GPL will come up with an update to address it, similar to how SaaS forced GPL to update to AGPL.

[Libs]

7 hours ago, Buckeye said:

Interesting article in this weekend’s WSJ from Christopher Mims titled, “The AI Superstars Who Say a ‘Vibe Slop’ Crisis is Coming.”  I’m surprised it hasn’t gotten more attention with regards to the Saaspocalypse.  

 

The article starts….”Two engineers who built the core of the massively popular OpenClaw AI agent have a stark warning: The artificial intelligence supposedly capable of replacing well-paid software developers is flooding the world with bad, potentially even dangerous, code.” 
 

“You have infrastructure that’s falling apart, and you have software that’s now very, very buggy compared to before,” says Mario Zechner, creator of Pi, the agentic harness inside OpenClaw. “We can play this game for a couple more months, or maybe even years, but eventually it will catch up to us.”

 

The article concludes with…”Zechner believes a reckoning is coming. He thinks big companies will soon realize that their overemphasis on AI-produced code is driving up costs and leading to subpar software. ”

 

https://www.wsj.com/tech/ai/vibe-coding-slop-ai-tools-e6a99394?mod=author_content_page_1_pos_1

 

Here's something less encouraging, from a link in the article:

 

Raj Sharma, a global managing partner for growth and innovation at Ernst & Young, said the audit, consulting and tax firm doesn’t plan to do away with its longstanding ERP software from SAP, but it is vibe-coding and using AI agents to build its own customizations on top of it.

EY, which has an annual technology budget of $1 billion, is able to save some of those IT dollars by vibe-coding rather than purchasing upgrades directly from SAP, according to Sharma. 

“If this AI wasn’t there, vibe-coding wasn’t there, the agentic frameworks weren’t there, it would’ve taken another very costly, expensive upgrade of the SAP software,” he said.

Thimaya Subaiya, executive vice president of operations at Cisco Systems, said the networking-equipment maker replaced a presentation software tool with its own AI agent, saving the company nearly $5 million annually in license costs. Cisco is also looking to replace other software vendors—some of which are costing the company $50 million to $200 million each year in subscription fees—with AI-created tools, he said.

“You need to look at all the applications that you’re using and say, ‘Which one of these can become automated workflows where we don’t need this application anymore?’” Subaiya said.

Replacing a piece of business software involves replacing business processes with AI agents, Subaiya said, and those bots then become the interface by which employees interact with software. “You don’t need an application anymore, because an application just becomes part of the agent database,” he said

Seemantini Godbole, chief digital and information officer of Lowe’s, said the home-improvement giant needs to carefully spend its IT budget, and vibe-coding its own content-generation software is one way to do it. Lowe’s distributes a circular to every store, for instance, and it’s a task that regularly requires new image creation.

 

At the same time, some small and midsize companies have successfully vibe-coded their own CRMs. Business technology leaders say smaller firms don’t encounter the same level of human, regulatory and legal complexities as they do, and can more easily vibe-code CRM or ERP software to suit their needs.

Amjad Masad, founder and chief executive of Replit, said many of the vibe-coding startup’s customers are small and medium-size businesses that build their own CRMs. But larger enterprises don’t—and shouldn’t—vibe-code their own Salesforce-like software. “What they want to do is create their own processes, their own workflows, their own agents, their own automations on top of it,” he said.

Kempe, Grant Thornton’s CIO, sees a similar trend playing out where the role of core business software changes. Rather than functioning as the heart of how businesses operate, it shifts into becoming a source of corporate data. AI agents will lead that charge—becoming the means by which businesses interact with software, he said.

Godbole, the tech chief at Lowe’s, said it’s not likely that all employees will need a “full-fledged” CRM in the future, and more likely they’ll need a simpler, AI agent or chat-based interface for the software.

While there will be some software vendors who will be “extremely innovative” in that future, there are also others where Lowe’s will say, “Actually, we can do this much better with OpenAI out of the box,” she said.

 

 

[Paarslaars]

On 5/23/2026 at 5:21 PM, Spekulatius said:

Can someone point out a single software business (SaaS or VMS) that has been disrupted by vibecoding or AI agents so far? I follow a bunch of software co but I don’t know any where this is clear.

We are entering year 3 of the AI revolution so you expect to see something.

Stack Overflow is basically dead... 

[frommi]

12 hours ago, Libs said:

 

Here's something less encouraging, from a link in the article:

 

Raj Sharma, a global managing partner for growth and innovation at Ernst & Young, said the audit, consulting and tax firm doesn’t plan to do away with its longstanding ERP software from SAP, but it is vibe-coding and using AI agents to build its own customizations on top of it.

EY, which has an annual technology budget of $1 billion, is able to save some of those IT dollars by vibe-coding rather than purchasing upgrades directly from SAP, according to Sharma. 

“If this AI wasn’t there, vibe-coding wasn’t there, the agentic frameworks weren’t there, it would’ve taken another very costly, expensive upgrade of the SAP software,” he said.

Thimaya Subaiya, executive vice president of operations at Cisco Systems, said the networking-equipment maker replaced a presentation software tool with its own AI agent, saving the company nearly $5 million annually in license costs. Cisco is also looking to replace other software vendors—some of which are costing the company $50 million to $200 million each year in subscription fees—with AI-created tools, he said.

“You need to look at all the applications that you’re using and say, ‘Which one of these can become automated workflows where we don’t need this application anymore?’” Subaiya said.

Replacing a piece of business software involves replacing business processes with AI agents, Subaiya said, and those bots then become the interface by which employees interact with software. “You don’t need an application anymore, because an application just becomes part of the agent database,” he said

Seemantini Godbole, chief digital and information officer of Lowe’s, said the home-improvement giant needs to carefully spend its IT budget, and vibe-coding its own content-generation software is one way to do it. Lowe’s distributes a circular to every store, for instance, and it’s a task that regularly requires new image creation.

 

At the same time, some small and midsize companies have successfully vibe-coded their own CRMs. Business technology leaders say smaller firms don’t encounter the same level of human, regulatory and legal complexities as they do, and can more easily vibe-code CRM or ERP software to suit their needs.

Amjad Masad, founder and chief executive of Replit, said many of the vibe-coding startup’s customers are small and medium-size businesses that build their own CRMs. But larger enterprises don’t—and shouldn’t—vibe-code their own Salesforce-like software. “What they want to do is create their own processes, their own workflows, their own agents, their own automations on top of it,” he said.

Kempe, Grant Thornton’s CIO, sees a similar trend playing out where the role of core business software changes. Rather than functioning as the heart of how businesses operate, it shifts into becoming a source of corporate data. AI agents will lead that charge—becoming the means by which businesses interact with software, he said.

Godbole, the tech chief at Lowe’s, said it’s not likely that all employees will need a “full-fledged” CRM in the future, and more likely they’ll need a simpler, AI agent or chat-based interface for the software.

While there will be some software vendors who will be “extremely innovative” in that future, there are also others where Lowe’s will say, “Actually, we can do this much better with OpenAI out of the box,” she said.

 

 

SAP consultants are the most overpaid software engineers in the world, probably not a bad thing if that goes away because anybody can vibe code their SAP modules on top of the service. But that doesn't mean all enterprise software goes away. Btw. these were always customer specific implementations and i don't think even SAP made money of of this, but not sure. The bulk of the money has gone to freelance SAP consultants for this.

Edited May 27 by frommi

[Spekulatius]

15 hours ago, frommi said:

SAP consultants are the most overpaid software engineers in the world, probably not a bad thing if that goes away because anybody can vibe code their SAP modules on top of the service. But that doesn't mean all enterprise software goes away. Btw. these were always customer specific implementations and i don't think even SAP made money of of this, but not sure. The bulk of the money has gone to freelance SAP consultants for this.

I don‘t think that’s true though.  I have a friend from school who does just this  at SAP. It not quite that simple because you have to understand exactly the customer requirements  and also what’s going on underneath (workflows, other software ties) and many other things. The coding is only part of the process. Anyone who says one can just vibe code it probably hasn’t done it.

 

That said, they do use AI to help with programming but again that’s only part of the job and not the largest part at that.

Edited May 28 by Spekulatius

[Longnose]

Funny thing is. I've been trying to better understand just how vulnerable SaaS is by getting in and doing it myself. Ive now made 3 apps... (stopmotion video app, Bi-Lingual Chinese Storybook App, Snake game app) Its bizzarely easy.  Ive come to the conclusion that I will never again pay for a simple app that does something simple. Though the chinese storybook is getting more and more complex layering in more and more layers and the code gets more and more complex and AI for as smart and dumb as it is makes a change but the change screws all the lower layers.  and it ends up rewriting large portions of the code cuz it doensnt know how to make simple changes.  

 

So if you take this towards the SAP developer or a larger enterprise software with tons and tons of layers. AI does not destroy this easily. For these its an enabler.  The code is only as good as the engineer / project manager that can direct to AI effectively.  Critical thinking, planning, structure, and execution become more of the driver than the "code" itself.  People who can learn these layers of framework will be the valuable employees. The developer role changes now more into a PM who can read and understand the code and understand software architecture more than just a straight "coder" 

 

IMO - AI is just an enabler for all these mega softwares. They will keep the best of their developers and software architects and be able to use AI to scale features and clear bugs faster. 

 

Not to mention distribution of said software/product.

 

So wierdly im still bullish on AI as an enabler of SaaS even when i can see how easy it is to spool up new apps and software's. 

[nsx5200]

4 hours ago, Longnose said:

So wierdly im still bullish on AI as an enabler of SaaS even when i can see how easy it is to spool up new apps and software's. 

It's that kind of nuance that non-technical people don't get easily.  They would rather have a very simple narrative that AI will take over anything that requires "intelligence".  Very black and white.

 

For those non-technical project managers/finance analysts, a crash course in AI helps to tease out the possible from the (probably) impossible.

[frommi]

6 hours ago, Spekulatius said:

I don‘t think that’s true though.  I have a friend from school who does just this  at SAP. It not quite that simple because you have to understand exactly the customer requirements  and also what’s going on underneath (workflows, other software ties) and many other things. The coding is only part of the process. Anyone who says one can just vibe code it probably hasn’t done it.

 

That said, they do use AI to help with programming but again that’s only part of the job and not the largest part at that.

Maybe i am just biased because in germany SAP consultants earn a lot more than normal developers. And if the software is just for 1 client, doesnt it make more sense that the client does the module himself? I doubt these things are super complex on the implementation layer.
I am sure the employees at the client doing it are much cheaper than hired consultants.

Edited May 28 by frommi

[Sweet]

8 hours ago, Longnose said:

Funny thing is. I've been trying to better understand just how vulnerable SaaS is by getting in and doing it myself. Ive now made 3 apps... (stopmotion video app, Bi-Lingual Chinese Storybook App, Snake game app) Its bizzarely easy.  Ive come to the conclusion that I will never again pay for a simple app that does something simple. Though the chinese storybook is getting more and more complex layering in more and more layers and the code gets more and more complex and AI for as smart and dumb as it is makes a change but the change screws all the lower layers.  and it ends up rewriting large portions of the code cuz it doensnt know how to make simple changes.  

 

So if you take this towards the SAP developer or a larger enterprise software with tons and tons of layers. AI does not destroy this easily. For these its an enabler.  The code is only as good as the engineer / project manager that can direct to AI effectively.  Critical thinking, planning, structure, and execution become more of the driver than the "code" itself.  People who can learn these layers of framework will be the valuable employees. The developer role changes now more into a PM who can read and understand the code and understand software architecture more than just a straight "coder" 

 

IMO - AI is just an enabler for all these mega softwares. They will keep the best of their developers and software architects and be able to use AI to scale features and clear bugs faster. 

 

Not to mention distribution of said software/product.

 

So wierdly im still bullish on AI as an enabler of SaaS even when i can see how easy it is to spool up new apps and software's. 

I think it’s going to be both good and bad for SaaS, I think it will depend on the product and the adaptability of the company itself.

[Spekulatius]

17 hours ago, Sweet said:

I think it’s going to be both good and bad for SaaS, I think it will depend on the product and the adaptability of the company itself.

I agree. It’s a new tool that leads to disruption and management as well as the domain makes a lot of difference here.

[MungerWunger]

https://www.reuters.com/business/anthropic-roll-out-claude-mythos-coming-weeks-launches-opus-48-2026-05-28/

 

few months ago: MYTHOS IS TOO DANGEROUS TO RELEASE

now: Anthropic to roll out Claude Mythos in coming weeks, launches Opus 4.8

 

If mythos were truly a big leap in capabilities, you'd think anthropic would keep it in-house to build much faster than their competitors

[kab60]

I stole this one from Twitter, as I had to try it myself, but I can't see how this doesn't end in a bust of some sort:

 

[backtothebeach]

1 hour ago, kab60 said:

I stole this one from Twitter, as I had to try it myself, but I can't see how this doesn't end in a bust of some sort:

 

Lol. There is no real intelligence in LLMs.

 

Perplexity has learned from this prompt being posted on social networks and gets it right: 

 

"Drive there. The car needs to be physically at the car wash to get cleaned, so walking would leave it at home. ..."

 

ChatGPT however:

 

"If it’s only 100 meters and the weather’s nice, walking is probably the better move. It’ll take about 1–2 minutes on foot, you avoid the hassle of starting the car just to move it a very short distance, and you get a bit of fresh air in the sun."

[thowed]

This is an interesting one - I don't like Altman, but I'm with ChatGPT on this one. 

 

The prompt doesn't specify that they are going to the car wash to wash the car.  Maybe they're meeting a friend there.

[patience_and_focus]

3 hours ago, thowed said:

This is an interesting one - I don't like Altman, but I'm with ChatGPT on this one. 

 

The prompt doesn't specify that they are going to the car wash to wash the car.  Maybe they're meeting a friend there.

I think there is something called priors. It is unlikely that you are going to a car wash just to meet a friend. It may happen in the real world but happens very rarely. An overwhelming number of times, trip to car wash is made only to wash your car. We make these default assumptions based on our priors (biases) all the time and answer rapidly. 

 

At least LLM should have done the (next?) best thing (if it didn't want to assume anything) which is to ask a question - "Is you intent to get your car washed?"

Edited Friday at 05:05 PM by patience_and_focus

[yesman182]

WTF is going on in software today? Away from my computer for 1 day and everything pops 5-10%?!

[roundball100]

11 minutes ago, yesman182 said:

WTF is going on in software today? Away from my computer for 1 day and everything pops 5-10%?!

Please let us know when you next plan to step away from your computer again ...

[Spooky]

Feels like the narrative is shifting. Becky from HR isn't vibecoding... 

[thowed]

NOW...