The AI crash is impossible - Change my View

[Marten98]

11 hours ago, Leo Gura said:

Almost zero.

@Leo Gura didn‘t you say a couple weeks ago that your strategy was to stack gold, since this is a foolproof way of investing, especially when stocks are down? And that gold is fairly cheap right now.

Are you keeping your money in cash or doing safe investments with lower interests?

I got a lot of money from selling my company right now and I want to consider different perspectives regarding the market right now.

[something_else]

18 hours ago, vinc3nc said:

@something_else I first invested in BTC in August 2017 at 4K. By the end of December (early January), I had 4x’d my portfolio in unrealized gains. That was probably the craziest bull market ever. Eventually, I lost 90% of it by not selling anything, partly because of inexperience and because it was my first crypto cycle.

In the second cycle (2018–2021), I barely invested any new money, thinking my altcoins would recover and that I’d make all the money back. Some of them did recover, but most didnt, so I failed my second crypto cycle as well. That was also when I got into spirituality, so I had less motivation to make money and pursue all this material stuff.

In 2022, I was like, “Enough is enough. If I don’t make it in this next cycle, I won’t be able to forgive myself.” So I locked in, got a job, and worked for two years, investing 80% of my salary every month into BTC and Solana at the lows. This time, I didn’t fuck it up. I made tons of profit, largely thanks to Solana going from 8usd to almost 300usd, and now Im well positioned for the next cycle. Let’s just say Im almost retired because of it.

The reason I had so much conviction is that I was staring at the screens all the time, studying the charts, listening to tons of people talk about investing and crypto, and being on X every single day, following the right people. Once you’ve done all that, there’s no doubt in your mind that BTC will always recover in the next cycle, no matter how hopeless it looks. But yeah, it takes a lot of effort and time to figure out which altcoins to invest in, because if you don’t, you'll get screwed. Every consecutive cycle gets harder and harder to make it (BTC ain't 4k anymore haha).

I'm going to be honest, this still just sounds like gambling psychology to me. Especially this part:

Quote

Once you’ve done all that, there’s no doubt in your mind that BTC will always recover in the next cycle, no matter how hopeless it looks

You never know when the cycle will be, what the highs will be or what the lows will be. None of these things are very predictable, which makes it much more akin to gambling than investing.

It's very analogous to the "just one more roll" mindset that you see from a gambler.

Also this confuses me:

Quote

I first invested in BTC in August 2017 at 4K. By the end of December (early January), I had 4x’d my portfolio in unrealized gains. That was probably the craziest bull market ever. Eventually, I lost 90% of it by not selling anything, partly because of inexperience and because it was my first crypto cycle.

You got in very early, if you'd sold nothing that bitcoin would be worth a fortune now, right? How did you lose 90% by not selling?

[vinc3nc]

 

@something_else So any investment that’s not super safe and predictable is in your opinion gambling? BTC is very volatile compared to other things, especially when it was a smaller asset. But if you study previous cycles and drawdowns, you can pretty much “know” what to expect in the next one. If there were suddenly some kind of fundamental issue with BTC, then I wouldn’t hesitate to reconsider things. But with BTC, nothing has ever fundamentally changed. Just because the price drops a lot doesn’t mean something is now broken.

The reason I lost 90% was because I had bought altcoins with the BTC I had. They went up way more than BTC, but they also crashed WAY more when BTC dropped to 3K in 2018. With alts, you’re definitely gambling. I wouldn’t say the same for BTC imo. It’s riskier not to own any.

To you, buying things after they’ve dropped 60%, 70%, 80%, or more might seem like gambling, but to me that has always been the best opportunity to make money. Maybe it’s because you don’t believe in BTC, or maybe you’re not aware of how much money is being printed all the fucking time.

What one calls gambling, another calls conviction. I still stand by this.

Edited yesterday at 12:56 PM by vinc3nc

[something_else]

2 hours ago, vinc3nc said:

So any investment that’s not super safe and predictable is in your opinion gambling?

I mean kind of, yea. It depends on the degree of unpredictability and the risk to reward ratio. IMO crypto blurs the line between investment and gambling.

In general I'd label anything that promises the possibility of a big win in a short timespan with a high degree of uncertainty/luck as at least gambling adjacent. This is the mechanism that creates gambling addicts.

Most altcoins and all memecoins are so unpredictable that they are just straight up gambling. BTC is better, but there is still an incredible amount of uncertainty.

Quote

But if you study previous cycles and drawdowns, you can pretty much “know” what to expect in the next one

I am quite sceptical of this. As the saying goes "past performance is not indicative of future results". There have been 3 cycles of four years, but that fact alone does not mean there is going to be a fourth. 3 is not a lot of datapoints either.

We are also at an incredibly uncertain point in the world with an unstable president of the US and the looming impact of AI. And there's a dynamic whereby the more people that believe the fourth cycle is going to be a money making opportunity, the less of a money making opportunity it becomes. All of these, among many other factors, compound the uncertainty.

Quote

Maybe it’s because you don’t believe in BTC

I believe in Bitcoin as a concept, but I would not personally rely on that belief as evidence that the price of Bitcoin is going to keep going up, or follow predictable patterns.

[vinc3nc]

 

@something_else Not 3. There have been at least 4 market cycles (4 halvings), not counting what happened from 2009 to like 2012,

The historical drawdowns have been 93%, then 87%, then 84%, then 77.5%, and now 54% in 2026. At the lows, BTC has always been doubted and even declared dead. But time and time again, it has shown its resilience and recovered to new ATHs. That’s why I wouldn’t bet against BTC.

Also, I might be more aggressive regarding my goals when it comes to investing. I’m not satisfied with 15–20% per year. I dont just want to preserve my capital and slowly grow it, only to become wealthy when I’m 55 or 65. If thats what you want to do, thats perfectly fine. And for most people, it’s not a bad strategy. Most people don’t even invest, so if you’re making 15–20% per year, you’re already a winner in a sense. My goal is to become financially free before I turn 40, or in my early 40s at the latest (not from 0, I’m already like 70% there).

Also, keep in mind that conviction doesn’t require certainty. It requires believing strongly enough to act despite uncertainty.

[something_else]

3 minutes ago, vinc3nc said:

But time and time again

4 times is not time and time again, the 5th cycle may exist under a completely different market scenario which does not produce the same reliable pattern.

4 minutes ago, vinc3nc said:

Also, I might be more aggressive regarding my goals when it comes to investing. I’m not satisfied with 15–20% per year. I dont just want to preserve my capital and slowly grow it, only to become wealthy when I’m 55 or 65. If thats what you want to do, thats perfectly fine. And for most people, it’s not a bad strategy. Most people don’t even invest, so if you’re making 15–20% per year, you’re already a winner in a sense. My goal is to become financially free before I turn 40, or in my early 40s at the latest (not from 0, I’m already like 70% there).

That's a fair goal, but in order to stand any chance of making 50%, 100% or more returns through only investing, you must accept a massive risk. You basically have to go all in, and if that's within your risk tolerance then fair enough I guess.

9 minutes ago, vinc3nc said:

Also, keep in mind that conviction doesn’t require certainty. It requires believing strongly enough to act despite uncertainty.

The market doesn't care about what you need or what you believe, though.

10 minutes ago, vinc3nc said:

My goal is to become financially free before I turn 40, or in my early 40s at the latest (not from 0, I’m already like 70% there).

That may actually be an argument to take on less risk. It's pretty easy to get from 70% to 100% of your goal using less risky investments. If getting to 100% quickly requires risking the 70% of your goal you've already achieved, that really is just gambling.

[vinc3nc]

@something_else BTC is only 17 years old. You can’t ask for a more reliable pattern.

For me, BTC = less risk, or low risk, and alts are on the other side of the spectrum. Other assets aren’t of interest to me at the moment because I think there’s an enormous opportunity in crypto right now, and it’s going to reprice violently to the upside.

Time will show whether I was right or not, and I have a feeling it will be the same as always: ‘You couldn’t have known. You just gambled and got lucky.’”

[Elliott]

Do any of you use any crypto for commerce? Buying and selling goods, not just trading crypto like beanie babies?

The only inherent value in crypto is from money laundering, and that's only until quantum computers start mining. It's like buying shares in shady laundromats. Consider if a fiat currency collapses, why would we revert to btc at any recent price?

Edited 23 hours ago by Elliott

[something_else]

24 minutes ago, vinc3nc said:

@something_else BTC is only 17 years old. You can’t ask for a more reliable pattern.

For me, BTC = less risk, or low risk, and alts are on the other side of the spectrum. Other assets aren’t of interest to me at the moment because I think there’s an enormous opportunity in crypto right now, and it’s going to reprice violently to the upside.

Time will show whether I was right or not, and I have a feeling it will be the same as always: ‘You couldn’t have known. You just gambled and got lucky.’”

I mean I wish you luck. Out of curiosity, what will you lose if you’re wrong?

[vinc3nc]

 

@something_else Depends on what “being wrong” looks like exactly. There are many levels to that, I guess. One is underperformance, another is slowly going to zero, and so on.

Anyway, it’s interesting to see the opinions of people outside the crypto bubble. Spending so much time on X and watching crypto YouTube can make you blind to the fact that you’re only listening to people who are in the bubble with you, which can make you overlook certain things.

So in general, crypto is still perceived very negatively by most people, and rightfully so, because of all the rugs and extraction that have happened. But on the other hand, it also makes me realize that we might still be early, because so few people have any knowledge of or interest in it, or simply don’t get it. Maybe we’re not that late yet.

[LordFall]

Gemini gave an 80% of some sort of altseason in the next 6 months due to the short term liquidity pump in M2 so I put a bit of money into some altcoins. I'll watch how it moves over the next few months and let you guys know. I bought some Pengu and Bittensor. 

It's not much money and I still think focusing on business and primary cashflow is more important but I think with short term liquidity this move could pan out. 

Edited 20 hours ago by LordFall

[oOo]

Anyone with genuine capability in AI infrastructure, model-building and entrepreneurship should prepare for a coming shift, intelligence becoming privately owned infrastructure. That's because as GPUs cheapen and open models improve, expect an industry resembling solar installation, except technicians install home, business and community AI systems. All that being locally hosted, internet-connected, private and continuously specialised around their owners.

And this isn't a few years down the track, this is starting by the end of this year, 2026. Something that took millions of dollars to do, will only cost low thousands in the very fastly approaching future. 

If you like electronics, especially if you're say an electrician (you don't need to be a genius in engineering) and you're wondering what kind of AI company you could build that moulds the worlds together, this is precisely what's coming less than a year away given the costs will be going way down. If I were wearing those shoes, that's precisely what I'd be doing, it'll become incredibly in high demand, and you'll be able to spend time coming up with all the unique ways it'll apply before the playbook has been written and then market it to the people that don't know they even want it yet. I would begin immediately planning just how to start a business around this inside of 12 months the latest, it'll be a gold rush in this specific area of the industry alone.

That changes the strategic position of OpenAI, Claude and other cloud providers. Something that's going to translate into a future from “everyone rents intelligence from a few companies", rather that general intelligence becomes commodity infrastructure while specialised intelligence becomes locally owned capital. Security-sensitive firms, families and communities will especially value systems whose memory, models and developmental feedback loops remain sovereign. I'm already prepping to have my home hooked with all the necessary AI infrastructure that's all purpose built, and its not something that's even totally out there. A middle to high income earner with the right chops could start planning for it now, when I am a father I certainly don't want my kids using company based AI. It'll actually be amazing, because I'll be able to take what I know about teaching, learning and feedback loops into creative methodologies specifically built to teach my kids how to engage with the digital world and the greater world around them. This, is just an incredibly leap we're making outside the big AI companies. 

And this is still only conventional silicon AI. Biological and hybrid architectures are another frontier entirely. And if you're worried about robots taking over the employment base that's still at least 7-10 years away. 

As for an "AI crash", well, likely not possible. As for company crashes around big AI (I.e. Claude, OpenAI), well based on this alone that's far, far more likely and arguably very plausible in light of this reality alone, which will be the impetus for all the other realities that will be generated following this inspiring cornerstone.

Edited 11 hours ago by oOo

[Elliott]

1 minute ago, oOo said:

Anyone with genuine capability in AI infrastructure, model-building and entrepreneurship should prepare for a coming shift, intelligence becoming privately owned infrastructure. That's because as GPUs cheapen and open models improve, expect an industry resembling solar installation, except technicians install home, business and community AI systems. All that being locally hosted, internet-connected, private and continuously specialised around their owners.

And this isn't a few years down the track, this is starting by the end of this year, 2026.

If you like electronics, especially if you're say an electrician and you're wondering what kind of AI company you could build that moulds the worlds together, this is precisely what's coming less than a year away given the costs will be going way down. If I were wearing those shoes, that's precisely what I'd be doing, it'll become incredibly in high demand, and you'll be able to spend time coming up with all the unique ways it'll apply before the playbook has been written and then market it to the people that don't know they even want it yet. 

That changes the strategic position of OpenAI, Claude and other cloud providers. Something that's going to translate into a future from “everyone rents intelligence from a few companies", rather that general intelligence becomes commodity infrastructure while specialised intelligence becomes locally owned capital. Security-sensitive firms, families and communities will especially value systems whose memory, models and developmental feedback loops remain sovereign. I'm already prepping to have my home hooked with all the necessary AI infrastructure that's all purpose built, and its not something that's even totally out there. A middle to high income earner with the right chops could start planning for it now, when I am a father I certainly don't want my kids using company based AI. It'll actually be amazing, because I'll be able to take what I know about teaching, learning and feedback loops into creative methodologies specifically built to teach my kids how to engage with the digital world and the greater world around them. This, is just an incredibly leap we're making outside the big AI companies. 

And this is still only conventional silicon AI. Biological and hybrid architectures are another frontier entirely. And if you're worried about robots taking over the employment base that's still at least 7-10 years away. 

As for an "AI crash", well, likely not possible. As for company crashes around AI, well based on this alone that's far, far more likely and arguably very plausible in light of this reality alone, which will be the impetus for all the other realities that will be generated following this inspiring cornerstone. 

This isn't true though, AI needs massive datasets to work. When you hear about private AI chips, they mean for businesses, like Walmart, it would still be an entire building.

[oOo]

5 minutes ago, Elliott said:

This isn't true though, AI needs massive datasets to work. When you hear about private AI chips, they mean for businesses, like Walmart, it would still be an entire building.

You’re conflating training AI with running AI.

Training frontier models needs massive datasets and data centres. Running already-trained open models does not. Today, 20B–120B models can run locally on workstation-class hardware. Meaning, personalisation comes from memory, retrieval and just fine-tuning, not rebuilding GPT from scratch. 

So different problem with radically different infrastructure. Haha, honest misinterpretation all good.

[oOo]

Just to add as well @Elliott , your phrases “AI needs massive datasets to work,” “businesses, like Walmart,” and “entire building” assume knowledge will keep being carried around like truckloads of books man. I think it becomes more like electricity or a phone call, so stored where useful, compressed where possible, streamed when needed.

Rule of thumb now: information gets smaller, faster, cheaper and more connected overtime. That's not a pattern anymore, that's becoming law, and its around the corner for larger and larger datasets for the private user. 

Your phone doesn’t contain everyone’s voice for example in order to transfer it from one phone to the next; the network briefly holds and moves the right signal at the right moment. AI can increasingly work the same way like a private brain at home, your personal memory beside it, giant libraries elsewhere and only the needed knowledge flowing in with specialised feedback loops customised for you and designed by the future architects of informational tomorrow. That's a future job description for workers, a future company for the entrepreneur, and a future skill-set I'm honing right now. 

Also, one more word to add to that "Rule of thumb" is *criticality*. Meaning, a fucking library now becomes something your AI can continuously question, rearrange and learn from, not shelves you manually search, this turns barcode into evolving DNA Imagine what that means on the relationship between patterning human behaviours and upgraded choice metrics. 

And by the way, the specific cost is already wild, NVIDIA’s desk-sized DGX Spark is US$4,699, with 128GB memory and support for models around 200B parameters. That's fucking raw dice.

“Walmart” might actually be the boring customer just to go back to your example there as well. Imagine a small importer owning an AI that permanently learns suppliers, shipping, exchange rates, customs, competitors and product demand, and that's rather than staff becoming exhausted professional ChatGPT prompters heh!

And on top of that let's gently take a zoom out going back to my earlier point, that is, if private intelligence gets cheaper, the roads carrying that intelligence get cheaper too, meaning closer and closer to our own micro version of the global internet. This means where brain, library and network are all shrinking toward private ownership, more and more communities will pool them and centralised government will find it harder and harder to win people over through a rigged two party system. This means that cults may also increase in the future as well haha, just when you think they're decreasing overtime. 

Also one more divergent link is on neurotech following later, but just only as hypothesis at this point, that's going to more and more entre the hands of the individual as well. And that's because the present hierarchical scaffolding between research and end consumer is going to level far more  as well where anyone with the intelligence and raw compute, is going to have to worry less about losing their dignity over begging for an investor grant where the loop between their own research, experimentation and breakthroughs  becomes tighter and tighter. 

This reminds me of a guy by the named of Stephan Molyneux, ha! I know some people may sigh when I mention his name but his mindscape rattles well here concerning applied anarcho-capitalism following this intelligence revolution, while simultaneously being a working hypothesis for decentralised Jacque Fresco "Venus Project" (research the man and the term/name). Blending the two worlds, some with the coming new fragrance of cult smog others far less, and will be far cooler as an individual, family and even local community.

That's pretty fucking wild. 

I think that military school may just be the coming worst place to send your kids in the coming future though but haha; comparatively, albeit bad in some ways throughout all time, its strengths will become weaker and weaker in light of AI undoubtedly being used to monitor how many butt farts a student does in a day and the corresponding disciplinary action that should follow. Man, that's wretched! I actually do wonder how they will respond to this revolution other than say better brainwashing recruits on their coming role to protect planet earth from future space wars ha!

 

Edited 9 hours ago by oOo

[oOo]

35 minutes ago, oOo said:

Just to add as well @Elliott , your phrases “AI needs massive datasets to work,” “businesses, like Walmart,” and “entire building” assume knowledge will keep being carried around like truckloads of books man. I think it becomes more like electricity or a phone call, so stored where useful, compressed where possible, streamed when needed.

Rule of thumb now: information gets smaller, faster, cheaper and more connected overtime. That's not a pattern anymore, that's becoming law, and its around the corner for larger and larger datasets for the private user. 

Your phone doesn’t contain everyone’s voice for example in order to transfer it from one phone to the next; the network briefly holds and moves the right signal at the right moment. AI can increasingly work the same way like a private brain at home, your personal memory beside it, giant libraries elsewhere and only the needed knowledge flowing in with specialised feedback loops customised for you and designed by the future architects of informational tomorrow. That's a future job description for workers, a future company for the entrepreneur, and a future skill-set I'm honing right now. 

Also, one more word to add to that "Rule of thumb" is *criticality*. Meaning, a fucking library now becomes something your AI can continuously question, rearrange and learn from, not shelves you manually search, this turns barcode into evolving DNA Imagine what that means on the relationship between patterning human behaviours and upgraded choice metrics. 

And by the way, the specific cost is already wild, NVIDIA’s desk-sized DGX Spark is US$4,699, with 128GB memory and support for models around 200B parameters. That's fucking raw dice.

“Walmart” might actually be the boring customer just to go back to your example there as well. Imagine a small importer owning an AI that permanently learns suppliers, shipping, exchange rates, customs, competitors and product demand, and that's rather than staff becoming exhausted professional ChatGPT prompters heh!

And on top of that let's gently take a zoom out going back to my earlier point, that is, if private intelligence gets cheaper, the roads carrying that intelligence get cheaper too, meaning closer and closer to our own micro version of the global internet. This means where brain, library and network are all shrinking toward private ownership, more and more communities will pool them and centralised government will find it harder and harder to win people over through a rigged two party system. This means that cults may also increase in the future as well haha, just when you think they're decreasing overtime. 

Also one more divergent link is on neurotech following later, but just only as hypothesis at this point, that's going to more and more entre the hands of the individual as well. And that's because the present hierarchical scaffolding between research and end consumer is going to level far more  as well where anyone with the intelligence and raw compute, is going to have to worry less about losing their dignity over begging for an investor grant where the loop between their own research, experimentation and breakthroughs  becomes tighter and tighter. 

This reminds me of a guy by the named of Stephan Molyneux, ha! I know some people may sigh when I mention his name but his mindscape rattles well here concerning applied anarcho-capitalism following this intelligence revolution, while simultaneously being a working hypothesis for decentralised Jacque Fresco "Venus Project" (research the man and the term/name). Blending the two worlds, some with the coming new fragrance of cult smog others far less, and will be far cooler as an individual, family and even local community.

That's pretty fucking wild. 

I think that military school may just be the coming worst place to send your kids in the coming future though but haha; comparatively, albeit bad in some ways throughout all time, its strengths will become weaker and weaker in light of AI undoubtedly being used to monitor how many butt farts a student does in a day and the corresponding disciplinary action that should follow. Man, that's wretched! I actually do wonder how they will respond to this revolution other than say better brainwashing recruits on their coming role to protect planet earth from future space wars ha!

 

 

Just for some humour regarding what I said directly above.

Ha, no one will expect this.

I always listened with an open minded, never really judging.

It's amazing how we'll see anarcho come up in the way I've just given it the curtain call. 

The Molyneux factor

(read quoted text for context).

Predicted the future, unspoken AI political genius!

 

 

 

 

The Fresco Factor. 

I don't think looking back, any of us anticipated the mergence of these two men's work!

Good ol' Jacque, rest in peace sir.

 

 

 

 

Some people may dislike both men, but their merging is inevitable sorry haha.

It's a strain of ideological species that is now in its incubation phase of hatching; and its the only logical response to what will become a fallen centralisation rather than your 1984 Orwell doom fantasy. So cheer up! Just be more mindful of potential cults as I talk about there increase their too heh.

 

 

 

Edited 9 hours ago by oOo

[Elliott]

9 hours ago, oOo said:

You’re conflating training AI with running AI.

Training frontier models needs massive datasets and data centres. Running already-trained open models does not. Today, 20B–120B models can run locally on workstation-class hardware. Meaning, personalisation comes from memory, retrieval and just fine-tuning, not rebuilding GPT from scratch. 

So different problem with radically different infrastructure. Haha, honest misinterpretation all good.

That's not "AI", that's a program made by "AI". "AI" means using it on something novel which means needing all data, processing a TON of data.

Edited 2 hours ago by Elliott

[something_else]

43 minutes ago, Elliott said:

That's not "AI", that's a program made by "AI". "AI" means using it on something novel which means needing all data, processing a TON of data.

Ehm, no, sorry that's just not accurate.

The trained models are general purpose enough that they can be used for novel tasks, especially inside an agent harness like Claude Code or Codex. They are still AI.

[Elliott]

1 hour ago, something_else said:

Ehm, no, sorry that's just not accurate.

The trained models are general purpose enough that they can be used for novel tasks, especially inside an agent harness like Claude Code or Codex. They are still AI.

Without data? No. What you're referring to is a program created by AI. Have you ever used automatic tools in photoshop? Without massive infrastructure you have relatively little data, effectively spoonfed (programmed). Its a local small-language model. The smaller your infrastructure the more precise your program and dataset needs to be tailored to your use. You can't just download a llm on a local lab and ask it to create a spaceship, and then how to cure cancer, you program it for a specialized task and spoonfeed it filtered data. It will be a calculator, not a polymath ChatGPT. I'm not saying it's not useful, it will be essential for many businesses, these AI companies are stealing their data. And if you outsource your dataset your security is gone, so you're effectively paying more for less. Perhaps someone will build data centers with some sort of true VPN style access for you, that wouldn't be bulletproof but perhaps sufficient for hobbyists, but then why not also buy the cheaper cloud AI under the same premise?

Edited 36 minutes ago by Elliott

[Pudgey]

@KosmicGeng Thank you for sharing your story! It is quite interesting and hits several core points. 

Just today, a Chinese company Z.ai just launched an extremely cheap model (GLM 5.3 flash): $0.15 / $0.50 per 1M, running on Chinese servers and chips. It matches Opus 4.8 on benchmark, but 10x cheaper.  It was free for 7-days on Openrouter and everyone thought it was Google or a large US AI company. It ended up being GLM (ZAI). USA banning China on Nvidia chips didn't do anything but cause them to hyper-optimize their models, making them cheap and just as capable. 

I don't see how Anthropic and OpenAI survives if open weight models are becoming this good. Example: Anthropic Sonnet 5 ($2.00 per million input tokens and $10.00 per million output tokens) and Google 3.7 flash ($0.75 per million input tokens and $3.75 per million output tokens, but will increase to $1.50 per million input tokens and $7.50 per million output tokens starting January 1, 2027). Deepseek and GLM are by far cheaper for Enterprise companies to transition to, even hosting their own server rakes (AT&T is already doing that). 

Yet, here's China making a model better than Google for even cheaper, that will become open source for US inference companies. 

The AI crash is very possible (NOT IMPOSSIBLE) and it'll most likely happen to US AI companies that are evaluating themselves to be worth trillions (Example: Anthropic is preparing to tell IPO investors that it sees its total addressable market (TAM) exceeding $30 trillion):  yet, OpenAI and Anthropic has only solved coding, nothing else. LLMs have not proven anything outside of software development - maybe that'll change (or NOT). 

Google has accomplished complete dominance by embedding Gemini across multiple products for free, with a worse model than Anthropic and OpenAI. Enterprises are going to slowly transition to China tech (or other close alternatives) to avoid those big API bills. Once that transitioning starts to happen, that's when the bubble starts to pop. It won't happen to all AI companies, but OpenAI will be the first to go...pop. 

LLM prices are crashing to the bottom and it doesn't look like it's stopping.