AI vs. software companies

[LC]

3 hours ago, whatstheofficerproblem said:

As for who wins, obvious answer would be Verisk, Constellation, Thompson Reuters & Epic systems kind of companies that have siloed prop data that these model sellers won't have access to. No amount of intelligence artificial or otherwise can conjure up 40 years of financial terminal data or the medical records of half of America's hospitals. The data is the alpha.

 

What's stopping model producers from stealing that data? Or temporarily training on it? 

[whatstheofficerproblem]

1 hour ago, LC said:

What's stopping model producers from stealing that data? Or temporarily training on it? 

 

Sure, where will they steal it from?

[LC]

All these corporate end-users with proprietary data will be using a combination of models from the oligopoly of model producers (google, anthropic,openAI, etc.). Those models will have access to that data, don't you think it's feasible those producers will find some way to leverage/integrate that particular domain-specific data?

[adventurer]

16 hours ago, whatstheofficerproblem said:

 

The counter argument defeats itself. Yes, you're right on it being lowest hanging fruit as it's deterministic. But even in this "easiest" domain, METR studies show experienced developers are 19% slower with AI, AI-generated code produces 1.7x more bugs, code churn has nearly doubled, and 95% of GenAI pilots fail to reach production. If AI can't reliably close the good-to-great gap in the domain with the tightest, most unambiguous feedback loop available, what exactly is the basis for believing it will handle domains where feedback is messier, more subjective, and more delayed?

 

The AGI argument has a dependency chain, current models improve -> we get AGI -> AGI solves everything. Each step here requires clearing a bar higher than the last. My point is that probabilistic models are the wrong mathematical tool for modeling reality. You will never reach AGI using RL.

 

Discussion around AGI feels a lot like cold fusion to me lmao. "sure, the current business model doesn't work, but once we achieve cold fusion, energy will be free." Prediction of AGI as you mention above, is not evidence of it but rather a bet. Who are people making the loudest predictions here? Usual suspects that are spending hundreds of billions into this bottomless pit with the most financial incentive to make those predictions.

 

I mean, Sundar himself dodged the AGI bullet and even called it semantics. Great, so even Google, spending $180 billion in capex, won't explicitly commit to an AGI timeline.

 

You'd have to assume a monotonic curve i.e. capabilities keep improving, therefore they will keep improving until AGI. But the research once again shows something radically different. Positive sentiment toward AI tools dropped from over 70% to roughly 60% in 2025. Trust in AI-generated code accuracy fell from 40% to 29%. AI projects have an 80% failure rate that has remained stubbornly consistent despite better tools and growing expertise. These are not the metrics of a technology on an exponential path to transcendence but looks like something hitting a ceiling and plateauing to me.

 

 

I allude to this in my write-up above. I'm not arguing AI isn't transformative, but we need to stop conflating technological potential with economic accrual. Fiber was important and transformative, who made money? Cloud was important and transformative, who made money?

 

The question never was if AI will be important, but who will make the money. Even if AGI arrived tomorrow, the economic questions we've been discussing don't disappear. They get worse. If AGI can do everything, then the model is the ultimate commodity as in everyone will have it, nobody can differentiate on it, and value accrues to whoever has the proprietary data, the embedded workflows, and the customer relationships. Which is, again, the software companies, not the model providers.

 

Think about what AGI actually means if you take the claim seriously. It means a system that can do any cognitive task at human level or above. Now follow that to its logical conclusion. If OpenAI builds AGI, what stops AGI from building a competitor to OpenAI? If the system is truly generally intelligent, it can design new architectures, optimize training runs, discover novel algorithms. The moment AGI exists, the intellectual moat a la "Saar we have the smartest engineers saar" evaporates because the AGI is the smartest researcher. And it's running on hardware that anyone with enough capital can rent. The scarcity was always human genius and doesn't AGI eliminate it?

 

The open-source dynamic accelerates this argument to absurdity. DeepSeek replicates frontier capability at a fraction of the cost using clever engineering. That clever engineering was done by humans who are scarce and expensive. In a post-AGI world, the clever engineering is done by the AGI itself, which means the replication cycle collapses from months to days or hours. Meta releases Avocado or Llama or whatever, AGI optimizes it overnight, and now everyone has frontier capability.

 

The model providers' lead time vs peers is their most valuable asset and moat and that would shrink to nothing. I said Google survives because it embeds AI and vertically integrates. But if AGI is generally available, every competitor can build equivalent products. The reason Google's integration works today is that it takes thousands of brilliant engineers years to build these systems. AGI removes that constraint and suddenly a startup can build a search engine, a video platform, a cloud infrastructure all in weeks. Google's moat wasn't really the AI, never was. It was the accumulated complexity that only a massive organization could manage. AGI is the universal solvent for accumulated complexity, at least that's what it inherently means.

 

Today, the stack looks like: hardware -> foundation model -> application -> customer. Value concentrates at the model layer because building models is hard. In a post-AGI world, building models is trivial as the AGI can do it. Building hardware is still constrained by physics as you need fabs, silicon, rare earth materials etc. So value either drops down to hardware and energy (things that are physically scarce and can't be replicated by intelligence alone) or rises up to the customer relationship layer (brand trust, regulatory approval, data access, contractual lock-in). The model layer, which is pure intelligence, becomes the cheapest part of the stack because intelligence is now abundant.

 

If you zoom out and look at all this retardation from a philosophical perspective, intelligence has never been the bottleneck for most economic activity. The bottleneck is trust, coordination, regulation, physical resources, and institutional legitimacy among other things. A brilliant doctor who isn't licensed can't practice medicine. A genius architect whose building isn't up to the building code gets shut down.  AGI adds infinite intelligence to a system where intelligence was already becoming abundant and other things were scarce. It's like adding infinite water to a desert where the constraint is actually arable soil.

 

So, AGI bull case is a paradox. The more powerful the intelligence becomes, the less the intelligence itself is worth, because it's the one thing that's no longer scarce. Everything else becomes relatively more valuable and what becomes valuable is not something these guys can provide.

 

That was all a mouthful, but yeah, been going down on this rabbit hole for while.

Thank you. How did you end up with these conclusions? Like the abundance of intelligence causing it to be almost worthless while its vast quantity boosts profitability of the already dominant Saars-companies. 

 

Most people are stuck at the part where intelligence starts to become abundant causing them to think that its exponential growth in quantity, so to speak, causes exponential growth to the moon and beyond, wiping out every competitor that does not own that intelligence. To my mind that sums up the current conversation about AI and its economic impact.   

Edited April 21 by adventurer

[whatstheofficerproblem]

13 hours ago, LC said:

All these corporate end-users with proprietary data will be using a combination of models from the oligopoly of model producers (google, anthropic,openAI, etc.). Those models will have access to that data, don't you think it's feasible those producers will find some way to leverage/integrate that particular domain-specific data?

 

LC, c'mon, they're obviously not that stupid. Enterprise contracts explicitly prohibit training on customer data, all the model providers have been forced to explicitly offer zero data retention options. While in no way do I trust them, if it ever leaked that they were secretly stealing that enterprise data, end result would be an extinction level event where you lose all of your enterprise revenue.

 

 Let's say these clowns try anyways right. Ok, if we take VRSK for example, their ISO division collects claims data from virtually every property and casualty insurer in the United States so they have like decades of granular claims history from fire to cars to house to everything which is reported by the insurer into their pool of data.

 

No single insurer has the full picture, only VRSK does as they are the trusted aggregators, they have become the central nervous system of the insurance industry effectively. So assume the model providers somehow get access to the raw data here, which would be next to impossible, but let's assume. Ok, you have the data, now monetize it. But how will they monetize it? You'd be missing the actuarial models, rating algorithms, loss cost calculations, and statistical plans built on top of decades of cross-referenced data. Insurance pricing in America literally runs through Verisk's statistical agents. State regulators approve rate filings based on Verisk's loss cost analyses. There is a whole regulatory infrastructure aspect to the pricing of insurance.

 

The co-operative structure they have going on, along with the data is the moat. Insurers contribute their data to Verisk because they get back something none of them could build alone in the form of industry-wide statistical credibility. A small regional insurer writing 5,000 home policies can't price wildfire risk accurately from their own experience. But Verisk aggregates across the entire market, giving that small insurer actuarial credibility it could never achieve independently.

 

The more insurers contribute, the better the models get, which attracts more insurers. OpenAI can't replicate this by being smart. It would need every insurer in America to voluntarily send them claims data, and there's zero reason for them to do so when the Verisk relationship already works. 

 

I can much deeper into the Verisk context, but that is a post for the VRSK thread at a later date. Your question is a game theory problem imo. Let's say OpenAI secretly trains on Goldman Sachs' data flowing through their API. What have they gained? A model that's slightly better at Goldman's specific use cases. But Goldman is one customer. JPMorgan's data is different. Citadel's data is different. The proprietary value is in each firm's specific, unique dataset i.e. their positions, their strategies, their client relationships. Aggregating across customers gives you generic industry knowledge, which is already in the training data from public sources. The alpha is in the specific, and you can't use firm-specific data without revealing to everyone that you're doing it.

 

These providers are competing with each other at the end of the day. The moment one provider is even suspected of training on customer data, the others will use it as a sales weapon and that competitive pressure creates a race to the bottom on data usage. Each provider is incentivized to be more restrictive than the others to win trust. If not it's very easy to just run the LLM locally for much of it's capability without risking your data getting stolen.

 

This is already happening imo. Companies are fine-tuning models on their own proprietary data and keeping the resulting weights in-house. Bloomberg built AskB which is very good actually, Epic is building AI into its own platform using its own data. They're using the foundation models as a base and adding their data as the differentiation layer on top. The value flows to the data owner, not to the model provider. The model provider gets the API fee. The data owner gets the competitive advantage.

 

So your argument has teeth in consumer data, but think enterprise data safe.

[whatstheofficerproblem]

3 hours ago, adventurer said:

Thank you. How did you end up with these conclusions? Like the abundance of intelligence causing it to be almost worthless while its vast quantity boosts profitability of the already dominant Saars-companies. 

 

Supply and demand issue. If every kid the US was say 180 IQ super genius then what would the value of that intelligence be in the country? It would just be the norm no?

 

The reason AI has this wow factor today is because it looks smart and gets stuff done fast (allegedly). So all the enterprise clients and everyone else is forking over API fees to try and use it and get as far ahead in their own competitive race as possible.

 

If every enterprise on the world has access to intelligence, i.e. the abundance of intelligence, what is the value of this intelligence? You're back to square one no? NOW will still be the same, ADBE will still be the same, WDAY will still be the same in terms of company and their prowess because they have access to same things all over again. It's like all of them hiring the same 1000 smart engineers, no piece on the chess board has moved.

[Milu]

2 hours ago, whatstheofficerproblem said:

If every enterprise on the world has access to intelligence, i.e. the abundance of intelligence, what is the value of this intelligence? You're back to square one no? NOW will still be the same, ADBE will still be the same, WDAY will still be the same in terms of company and their prowess because they have access to same things all over again. It's like all of them hiring the same 1000 smart engineers, no piece on the chess board has moved.

I’d look at it a bit differently. Previously the large enterprise companies you listed Adobe, servicenow, Workday, had the scale and budget to spend millions on human intelligence. In the old world this was hiring very smart developers and engineers. In the new world every other company now has access to the same intelligence, but now it is in the form of tokens.
 

I would say there is some degree of diminishing returns in that if you already have a large development team and loads of traditional resources then adding AI on top of that would improve things by a certain percentage, but for a start up looking to create a new workday solution and compete with them could catch up a lot faster than the incumbent can push ahead. 
 

Secondly for large end-customers who have their own massive development teams maybe they decide to build more custom solutions for their internal processes rather than buying off the shelf. Obviously the local golf course aren’t going to build a custom solution, but maybe Bank of America would, or maybe Google builds its own HR system rather than using workday, I believe they already have their own custom CRM solution. It’s not as black and white as it was before AI came along.

 

Im not smart enough to know how this all plays out but I see a lot of people on one side confidently predicting that AI will destroy all SAAS companies and a similar amount on the other side confidently predicting that there is now chance SAAS companies will get disrupted. 

[whatstheofficerproblem]

3 minutes ago, Milu said:

I would say there is some degree of diminishing returns in that if you already have a large development team and loads of traditional resources then adding AI on top of that would improve things by a certain percentage, but for a start up looking to create a new workday solution and compete with them could catch up a lot faster than the incumbent can push ahead. 
 

Secondly for large end-customers who have their own massive development teams maybe they decide to build more custom solutions for their internal processes rather than buying off the shelf. Obviously the local golf course aren’t going to build a custom solution, but maybe Bank of America would, or maybe Google builds its own HR system rather than using workday, I believe they already have their own custom CRM solution. It’s not as black and white as it was before AI came along.

 

Im not smart enough to know how this all plays out but I see a lot of people on one side confidently predicting that AI will destroy all SAAS companies and a similar amount on the other side confidently predicting that there is now chance SAAS companies will get disrupted. 

 

I alluded to this in my initial post. By the time start-ups gain the ability to rapidly catch up to the incumbents, the model providers won't stay solvent. Then again, Cloud computing was supposed to disrupt Workday & SAP, the mobile was supposed to disrupt Salesforce and open-source was supposed to disrupt Oracle. I think in each case startups did emerge and built a few incredible products at a higher velocity, but in every case the incumbents always adapted the same technology and extended their lead.

 

If you want to build another workday today, you'd be compiling compliance across 50 states, ERISA regulations, ACA reporting, multi-entity consolidation, and union contract variations where AI has no edge. What AI helps accel was never the hard part to begin with is my point. Some things have to take time and decades of customer feedback to correct course, there is not shortcut for that.

 

Your second argument actually predates AI. Large enterprises have always had the theoretical ability to build their own HR, CRM, and ERP systems. Google infamously uses myriad of internal tools. Think the reason most don't do it isn't the technical burden but opportunity cost. If we take Bank of America as example, their best engineers should be working on trading systems, fraud detection, quant strats and customer-facing products that generate revenue & not re-inventing payroll processing software.

 

This equation doesn't really change if SaaS vendors are similarly using AI to make their products better and cheaper, while building custom means your internal team now also needs to maintain, update, and secure an AI-augmented system indefinitely. Not to mention the amount of "mess" AI creates when using it to code.

 

I think there is a middle ground there in terms of AI enabling a new category of lightweight, purpose-built tools that don't try to replace SaaS incumbents entirely but pick off specific workflows where the incumbent is bloated or overpriced. So maybe we will see more of a death by thousand cuts instead of bullet to the head type of damage from AI.

 

Then again, I am skeptical because "Time" is of the essence.  In the short term, AI helps incumbents more than challengers because incumbents have the data, the customers, and the integration depth to deploy AI meaningfully imo. In the medium term start-ups, or challengers will use AI to attack the edges and we don't really know what happens long-term.

 

I'd say the model providers might not stay solvent starting medium term.

[LC]

7 hours ago, whatstheofficerproblem said:

This is already happening imo. Companies are fine-tuning models on their own proprietary data and keeping the resulting weights in-house. Bloomberg built AskB which is very good actually, Epic is building AI into its own platform using its own data. They're using the foundation models as a base and adding their data as the differentiation layer on top. The value flows to the data owner, not to the model provider. The model provider gets the API fee. The data owner gets the competitive advantage.

 

So I'll caveat the pushback I'm about to give with two things. And I'll also caveat that this is a little train-of-thought, just food for thought really:

 

1- I think end of day (and that's what matters) you're right - for the sole purpose that there are a TON of eyes on AI and I think the model providers are being very selective with their public perception. At least until they have more leverage.

 

2- My first real big boy job was in digital marketing, model development. The company I worked for "leveraged" data from every one of our clients without explicit consent. Anonymized it in a million ways and jumped thru all sorts of hoops to claim we weren't, but end of the day we were. So it does happen. It will especially happen if, as you say, tons of mini-LLM-providers come along without as much to loose as Google.

 

Now in terms of the pushback: Regarding enterprise: going from "Data is the moat" to "data and the 'co-operative structure' is the moat" is a bit of thesis drift, no? I work in a heaaaavily regulated area that is very similar to your insurance example above. The surrounding structure is easily replaceable. Actuarial models? They're in textbooks. The development data and input data feeding them, once either the model results, the presentations they are built around, the entire surrounding process flow of that "co-operative structure" - it's all visible to these LLMs and potentially the providers of those models. Same with a ton of enterprise customers. Executives and regulators are leaping in with both feet and blindfolds on. 

 

Then think about what happens in 3 years when these organizations, their systems and processes are all fully embedded leveraging the LLM compute of 3 model providers. What happens then? You've lost all your leverage to an oligopoly who will raise prices in a heartbeat and now you've got no choice because your talent is gutted and your 'proprietary data' is no more proprietary than your competitors. It really isn't. Vast majority of consumer and commercial financial products are incredibly similar. 

 

A lot of this is TBD and yet to play out but I just don't see a bunch of 100 year old companies who can barely manage their own data somehow have this come to jesus moment when they realize that they now hold all the power over Google who has been harvesting, managing, and training on data for decades. 

[Sweet]

@whatstheofficerproblem last 4-6 posts ITT have been the most realistic and clear headed views I’ve read on AI.  An antidote to the tech bros bullshit.  Very good stuff, thanks.

 

 

[whatstheofficerproblem]

@LC, the thesis drift was assuming they stole the data, anyways I'll indulge. Companies do find ways to launder data usage through anonymization, aggregation, and creative legal interpretation and I'm willing to bet Sama & Dario are already doing this. I'd reframe the initial drift as specificity. 

 

Lumping data is moat across the board with a blanket statement was a lazy way me trying to make the point, but moats still exist what I was trying to say. For Bloomberg it really is the data, Verisk it's the cooperative structure, Epic it's the workflow integration etc etc. To your point on actuarials and so on, I mean sure, but again how will they get the data? Occam's razor, I don't think the companies using these models are actually foolish enough to not confine it to a sandbox and limit access. 

 

The lock-in aspect of your counter argument is something in the back of mind constantly as of late. But once again, I'd think companies are not foolish given what AWS did to them. In AWS' case lock-in worked because they had a 5 year or so headstart and there were only three credible providers. The model space already has a ton of comps though.

 

OpenAI, Anthropic, Google, Meta's open-source, DeepSeek, Mistral etc. Just head to Ollama and you'll be assaulted by a large selection of models. Open-source here is viable in a way that open-source cloud infrastructure never was. If you fear lock-in, you can just run your choice of model locally. The oligopoly pricing scenario requires the open-source gap to remain large enough that enterprises can't credibly threaten to switch. From my point of view that gap is closing and enterprises retain leverage.

 

Google can build a better insurance pricing model. Google cannot get licensed to write insurance in 50 states, build relationships with 30,000 independent agents, and earn an AM Best rating overnight is my thinking.

 

Re: your point on the come to jesus moment for these old companies, I agree. But some individual companies within that category will still get destroyed by their own incompetence. That's always been true, with or without AI.  

[LC]

One last point I'll make before I am forced to go eat vegan pizza (maybe i'll be active on the "what are you drinking thread" later):

 

On the open source models: I don't think they will be a credible threat outside of perhaps code development. Look at FICO, they do nothing special nor have they for decades. The FICO 08 model is/was probably the highest volume model they have created, and it's a logistic regression model, for pete's sake. There have always been competitors with the same aggregated data and it takes not even a week to build a competitive model, but there is some lock-in there.

I think it will be the same for enterprise customers with Google, Anthropic, and OpenAI/MSFT. 

 

I agree with you Google is not going to start underwriting loans, but I don't think that Gemini/Vertex users are going to suddenly be able to squeeze Google.

 

I think it will be the other way around where current underwriting models don't really change but Gemini will be embedded in every other step of the insurance-writing ecosystem, and so Google will be able to squeeze (or more accurately, capture a percentage of headcount reductions) within that ecosystem. 

 

Put it this way: what happens when JPM replaces 1000 call center employees with a Gemini-based LLM that manages 95% of customer interaction? When Google goes to reprice that contract, who has the leverage? Even if JPM has Anthropic embedded in the same ecosystem, it's going to be oligopoly pricing from the model providers. JPM is not going to let Deepseek or Molt-whatever interact with their customers. And what happens when they go off Gemini and BofA doesn't? JPM is going to be more than happy saving 1000 salaries, they'll pay the model providers.

 

I mean yes, if these guys are smart they will do exactly what you mention and have 10 different LLMs tested, primed, and ready to manage that ecosystem, so they are not beholden to any one provider. But again I think their real win is the headcount cuts and not shopping around for model providers. 

[whatstheofficerproblem]

All good points. But FICO's power comes from the fact that Fannie Mae and Freddie Mac mandated its use for decades. It is literally a govt enforced monopoly. No model provider has anything close to that. Nobody is required by regulation to use Gemini or Claude. Every enterprise is choosing voluntarily, and can theoretically switch.

 

That enforcement led to the entire financial system being calibrated to their specific scale. Replacing FICO means recalibrating everything but the outputs wrt LLMs are very different, they are variable and not standardized. If JPM switches its call center from Gemini to Claude, the customers get slightly different phrasing at best. The downstream systems don't need to be rewired and recalibrated. There's no equivalent of "the entire securitization chain is built around this specific number."

 

When was the last time you used a wrench vs a swiss army knife? A wrench is harder to replace than a Swiss Army knife because the wrench fits the specific bolt perfectly. LLMs do everything, which paradoxically makes them more replaceable. The more general the tool, the more substitutes exist. FICO again had zero comp for decades while the model space already very crowded.

 

Switching from one LLM to another requires updating an API endpoint, adjusting some prompts, and running regression tests. It's not trivial, but it's not anaconda vice kind of lock-in either.

 

If I were to guess, these model providers would at best have more SaaS like pricing power i.e. 10-20% annual increases that annoy customers but don't trigger defection. Re: Labor math, I once again agree with you. But how many JPMs do you think exist in this world? There simply aren't enough enterprises that will spend the amount of money required to keep these monsters spending.

 

Let's say there are and these guys replaced enough labor and helped cut tremendous costs. But the toad has now tasted swan meat! The pandora's box is open. Model providers will start raising prices by how much ever the labor-replacement math supports. Enterprises now used to their new lush margins will start screaming. The industry associations will lobby. And the model providers will keep raising prices because the switching cost is astronomical. Just like FICO.

 

Then, five to ten years later, some combination of open-source models, regulatory intervention, and enterprise coalitions will force the door open to competition just like VantageScore is doing now. But by that point, the model providers will have extracted enormous value during the lock-in window. Sounds too good to be true.

 

Again, if JPM could save $50-70 million & Google charges $30M, that math is visible to every model provider simultaneously. Anthropic sees the same opportunity and offers to do it for $20 million. Microsoft undercuts at $15 million. The pricing anchor isn't the labor cost for long. It migrates to the competitive price almost immediately, because unlike FICO, there are multiple credible providers from day one. Labor replacement, while it sets a theoretical maximum from day 1, competition sets the actual price, and there's nothing structural preventing competition here.

 

Also, labor replacement is also a one-time event, not a recurring extraction. Once JPM fires  1,000 call center workers, they're gone. The model provider can't fire them again next year. The savings are captured once, and then the ongoing relationship is just an API contract that's subject to normal competitive dynamics. Google can't say "pay us more or we'll rehire your call center workers." The leverage disappears after the initial transition because the humans aren't coming back regardless of which model provider serves the contract.

 

You would also risk demand destruction here. If every bank replaces its call center with AI, the quality of customer service becomes undifferentiated. In a world where every bank runs the same Gemini-powered call center, wouldn't the bank running human powered centers become differentiated and command the premium? Commoditized service delivery leads to commoditized willingness to pay for it, which is unironically why call centers are cheap in the first place.

 

If we go into the philosophical side of the debate, this is principal-agent misalignment. The person making the AI purchasing decision isn't the person being replaced. It's the CFO or COO looking at headcount reduction & their incentive is to minimize the cost of the AI contract, not to maximize what they pay the model provider. Every dollar they don't pay Google flows to the bottom line as additional savings. So the enterprise buyer is highly motivated to play providers against each other, switch vendors, threaten open-source alternatives, and negotiate aggressively. This isn't a consumer market where people pay for convenience and enterprise procurement is the most ruthless buying environment in business.

 

Then we go back to singularity. If everyone is being replaced, no one has the money to spend, so at the cost of their margins, the companies firing and replacing with AI agents destroy each others' and their own customers which will kill their revenue stream. The whole economy goes to zero.

[whatstheofficerproblem]

1 hour ago, Sweet said:

An antidote to the tech bros bullshit.

 

I am a 'techbro' myself. Which is why I find the whole "AI saar" debate all the more appalling. It's wall street making the most noise followed by folks with vested interested via the model providers. Most tech employees see right through this.

[Spekulatius]

3 hours ago, LC said:

there are a TON of eyes on AI and I think the model providers are being very selective with their public perception. At least until they have more leverage.

Thats actually not correct. The AI companies are really bad public perception. If they were better, they would market their tools as work aides and smart assistants not as dooms day devices and mass job layoff enablers. AI is now feared by a lot of people and it’s only a matter of time until it creates optical pushback.

[LC]

Feared by employees, yes. Definitely not feared by executives and this administration. Both groups that make the decisions are pushing the technology quite hard and embracing the tech co's producing these models.

[LC]

Well, some of the producers (Trump vs. Anthropic, aside)

[Spekulatius]

40 minutes ago, LC said:

Well, some of the producers (Trump vs. Anthropic, aside)

I wonder if the Trump/ Hegseth still classify Anthropic as a supply chain risk, meaning neither Federal institutions (including military ) nor their suppliers or contractors can use it.

 

I guess Copilot and ChatGPT will have to do for the time being.

[whatstheofficerproblem]

@LC, I think we all know the guy in the office will make a deal with the devil if the devil was prepared to kiss his ring. There are no second order thoughts in the WH right now, but that will change. You're also severely underestimating "majoritarianism". A handful of executives won't stand a chance when the entirety of their workforce wants to push back.

 

@Spekulatius, they already classified it as such which is why Dario has been trying his best to kiss the ring. Palantir covering for them is the only reason there isn't a complete meltdown.

 

Interesting article.

 

https://blog.mozilla.org/en/firefox/ai-security-zero-day-vulnerabilities/

 

"A gap between machine-discoverable and human-discoverable bugs favors the attacker, who can concentrate many months of costly human effort to find a single bug"

 

"Encouragingly, we also haven’t seen any bugs that couldn’t have been found by an elite human researcher. Some commentators predict that future AI models will unearth entirely new forms of vulnerabilities that defy our current comprehension, but we don’t think so. Software like Firefox is designed in a modular way for humans to be able to reason about its correctness. It is complex, but not arbitrarily complex"

Edited April 22 by whatstheofficerproblem

[frommi]

If every software earnings call leads to IGV losing 5%, maybe some software stocks get to P/E's of 1 or 2 in 3-4 weeks?
Does someone understand whats so bad about NOW's report?

[whatstheofficerproblem]

36 minutes ago, frommi said:

Does someone understand whats so bad about NOW's report?

 

Basically Bill walked people off a cliff. Guided Adj. Op Margin down. You should keep in mind that beating consensus numbers doesn't amount to jack shit these days. You need to beat buyside's bogeys or you're a funding short. Revenue growth seen as not organic as it had acquisition tailwinds of like almost 200bps, then said Iran war headwinds lmao.

 

cRPO growing under 20% is not what people wanna see. Then you also add in the fact that he did an accelerated buyback at a much higher price when he knew the print was going to be bad, that has a lot of folks pissed and rightfully so.

 

"Don't fall for this parlor trick that a single button will replace 20+ years of software excellence" he said on the call. Also should know that he quoted WEB & Ben Graham with the whole weighing machine and voting machine analogy. That is a contra indicator in my eyes. PTSD from Sardar Biglari & Gary at Restoration Hardware.

[frommi]

I get that NOW is down -17% on GM misses and a tad high valuation for the uncertainty of terminal value, but why the rest of the software sector, that seems like very specific NOW problems? 
To me it also looks like the payments sector is attached to this as well, as if someone is liquidating a portfolio of high margin asset light businesses.

Edited April 23 by frommi

[whatstheofficerproblem]

21 minutes ago, frommi said:

I get that NOW is down -17% on GM misses and a tad high valuation for the uncertainty of terminal value, but why the rest of the software sector, that seems like very specific NOW problems? 

 

Welcome to the new age. IGV will continue to go down more whenever a software co misses than go up whenever one beats. It's all just flows at this point, this money is funding semis and semicap. If software is getting bid up, just know that it is degrossing and regrossing from a risk standpoint as positioning in semis is very overextended.

[frommi]

Is that some kind of pair trade, long semis, short software?
Will be interesting when that blows up.

Edited April 23 by frommi

[whatstheofficerproblem]

Yes, it's not a pair trade anymore. It's the ONLY trade you're allowed to make at most platforms which make up most of the market flows.