Leo's amazing predictions

[Leo Gura]

9 minutes ago, something_else said:

It is leaps and bounds above the AI image generation. Like not even in the same ballpark of impact.

That is because there is just so much more text as training data. The more training data, the easier it is to fake the illusion of AGI.

These tech companies are using the massive availability of intelligent text on the internet to create extremely convincing illusions of AGI that is fooling 90% of people.

But without AGI you cannot truly replace jobs, and what can be replaced becomes commoditized and almost worthless.

If you can just build a world-class app with a few prompts, then apps are now worthless. Say goodbye to apps-building as a business.

What people are not realizing is that whatever AI can do is no longer a viable business.

If I can code full self-driving cars with a few AI prompts, then Tesla self-driving is worthless.

The reality is that AI cannot build anything worldclass.

Edited yesterday at 11:19 AM by Leo Gura

[Davino]

2 minutes ago, Leo Gura said:

I don't think this is a valid argument because nuclear energy takes decades to build. You can't get significant new nuclear energy production without 5-10 years of build-out. This bubble cannot wait that long.

Yes that's the likely scenario but it still could be pulled off, we have been seeing major movements in this direction by Microsoft, Google, Amazon or Meta.

A great discovery in the energy field could start a new race that rapidly accelerates that process. But as you say the timing matters for this to work.

[Leo Gura]

6 minutes ago, Davino said:

Yes that's the likely scenario but it still could be pulled off, we have been seeing major movements in this direction by Microsoft, Google, Amazon or Meta.

A great discovery in the energy field could start a new race that rapidly accelerates that process. But as you say the timing matters for this to work.

But even if you have infinite free power, so what?

If AI cannot replace jobs then energy breakthroughs cannot change that.

The problem is going to be this: Companies will try to replace workers with AI, it will seem like it's working at first, but a few years down the road every company will realize that the products and services built via AI aren't worldclass, aren't competitive, making AI almost worthless. That's when the crash comes.

This is a debate about the inherent value of chatbots without AGI. Are chatbots without AGI worth trillions upons trillions? I don't think so. But I could be wrong about that.

Edited yesterday at 11:24 AM by Leo Gura

[Butters]

1 hour ago, Davino said:

If we get a breakthrough in nuclear energy, finding very efficient ways of producing massive amounts of energy, then keeping up the margin profits of building bigger AI models become more plausible.

But more compute will only lead to tiny improvements at this point, it's called the diminishing returns of AI scaling laws. At the current frontier, moving the needle just a fraction requires doubling or even 10x-ing the total compute budget. 

What actually makes sense is creating better harasses, which is why Claude Code and others have become significant in the last 12 months. 

[Leo Gura]

An important factor is China. China is making much cheaper models at 10% the cost.

China alone could kill OpenAI and Anthropic.

It will be a sweet irony if OpenAI and Anthropic both get replaced by cheaper foreign verisons that were copied off their work.

No honor among theives.

If you think about it from first principles, AGI should not use more energy than a single lightbulb. Since humans do it like that. But that means you don't need to pay Nvidia billions.

Edited yesterday at 12:34 PM by Leo Gura

[Davino]

1 hour ago, Leo Gura said:

But even if you have infinite free power, so what?

If AI cannot replace jobs then energy breakthroughs cannot change that.

Part of the problem companies are facing right now even when AI is able to replace part of a job, is that the cost associated with the use of the technology (tokens) is far greater than having a guy in India doing it. 

So the truth is that AI can't substitute most jobs but in those which it can, it's most of the times much cheaper to get a person to do it. A breakthrough in energy would mostly solve the second case. You would need a huge capex investment first to set up all the infrastructure and hardware but then the opex of running it would be much cheaper; making it probably affordable and sustainable.

[Leo Gura]

1 minute ago, Davino said:

Part of the problem companies are facing right now even when AI is able to replace part of a job, is that the cost associated with the use of the technology (tokens) is far greater than having a guy in India doing it. 

Token cost will go down a lot, I think. But AGI will not materialize so Indian jobs won't be replaced.

[Davino]

I'm excited to see how all this will unfold.

[Natasha Tori Maru]

I'm not so excited. I don't want the bubble to pop.

Not keen to experience another 2008.

[Leo Gura]

I am excited for mechanical hell. Hooray!

[Davino]

Don't be doomers, it's the cycle of life, let's enjoy it with popcorn 

[thenondualtankie]

4 hours ago, Leo Gura said:

These tech companies are using the massive availability of intelligent text on the internet to create extremely convincing illusions of AGI that is fooling 90% of people.

You're missing the fact that LLMs can generate their own text, verify whether this text is correct, and expand their own datasets.

This is why we're seeing such massive progress in verifiable domains like mathematics. AI can now solve research grade mathematics problems. You don't need to be "skeptical", look it up. There was a recent major theorem proven entirely by GPT 5.6-Sol.

 

Edited yesterday at 04:09 PM by thenondualtankie

[NewKidOnTheBlock]

If A.I. progress will be as slow as some of you are predicting and jobs will still be there, then that's also good as far as I'm concerned, it'd give me large enough of a time window to transition to a career in this field in some way, shape or form. Not seeing anything more worthwhile to be engaged with than with this tbh, as humans are just full of shit including everyone here and me (hence I'm opposed to the last Leo's blogpost about needing to invest into spectaculiar humans as opposed to A.I. LOL), I'm on the hype train until technology legit stops advancing with the pace it is advancing right now; economy and investors can kiss my ass; why should impatience of investors be indicative of anything other than their own foolishness

[Davino]

5 hours ago, Butters said:

But more compute will only lead to tiny improvements at this point, it's called the diminishing returns of AI scaling laws. At the current frontier, moving the needle just a fraction requires doubling or even 10x-ing the total compute budget. 

Yep, same as with Moore's Law and still haven't gotten around a better affordable solution since 1965 because we're so invested in silicon based transistors.

[Carl-Richard]

2 hours ago, thenondualtankie said:

You're missing the fact that LLMs can generate their own text, verify whether this text is correct, and expand their own datasets.

This is why we're seeing such massive progress in verifiable domains like mathematics. AI can now solve research grade mathematics problems. You don't need to be "skeptical", look it up. There was a recent major theorem proven entirely by GPT 5.6-Sol.

A wind-up toy can walk for a bit before it stops, yes.

[Joshe]

On 7/10/2026 at 6:14 AM, Leo Gura said:

But AI is not more intelligent than even a dumb human, even a teen.

AI is less intelligent than 100% of humans.

Notice that your logic assumes AGI. To be more intelligent than even the dumbest humans requires AGI.

Just because a chatbot piggybacks off 2000 years of human writing does not make it truly intelligent. It cannot operate without supervision and lots of correction. A chatbot is like a factory. It will work if you set it up perfectly, but this requires lots of effort and the tasks it does need to be very predictable and straightforward,

Why do you care if your problem solver has human-like intelligence or not if it can solve problems really well? 

AI is not some persistent always-on machine that exists in a cloud somewhere. You create instances of it and in those instances, it does a specific kind of work.

It's not designed to cut grass. But other actual dumb metal machines without AI are cutting grass and doing a fine job. The hunk of metal that cuts grass without human intelligence is solving a specific problem and no one gives a shit if it has intelligence so long as it gets the job done.

It would be like refusing to use a calculator because it doesn't understand math.

Edited yesterday at 07:24 PM by Joshe

[Joshe]

"Does it solve problem X?"

Even if all AI could do was summarize text and videos. How valuable is that one thing? Run it! 

Knowledge workers spend a lot of time reading docs, reports, email threads, meeting transcripts, tickets, articles, etc. If a tool reliably saves even 30% of time for each worker with regards to reading, how valuable do you think that is to collective humanity? Billions upon billions of man hours saved every year, and that's just solving "summarization". 

[Joshe]

"Just a chatbot" dissolves quick once you see Claude Code reliably run a 50-step process you've designed. For those who haven't experienced Claude Code at its full potential, you have no idea what this tech is capable of and no idea of it's value. 

I just had it setup a virtual machine on a different computer and I never even touched that computer. And on that VM, it setup desktop automation software to login to a specific website and take some code from my main machine and run it through a testing tool and then send the results back to my main machine. So now, I simply hit a hotkey to automate a routine 10-minute task that I'll never have to do again. This alone will save me 3-4 hours per month. And that's just ONE thing out of dozens I can streamline. That's about 42 hours a year from ONE automation.

I also set it up to so that it can access and troubleshoot my entire home network. Then I had it make all my lights, TVs, and all my IoT devices controllable from this keyboard I'm typing on. Then it setup a watchdog service on a Raspberry Pi that monitors all my devices and services and notifies me via a custom notification system I designed and it built if anything goes offline, and if so, sends notifications and reminders to specific devices depending on where I'm at. I then setup an email monitor service to watch for specific client emails that I need to see immediately, sent as a critical notification that forces my engagement, otherwise if they're not important, don't notify me but remind me about them later if I still haven't replied by EOD. I then had it build a secure, locally hosted password manager that's far easier to use than anything on the market.

It takes real design work and iteration to do these things but I did all this and much more in a weekend. It would have taken high-salary experts weeks or months to accomplish all of this without AI.

If you don't know shit about tech, you cannot fathom the value of this technology.

[Natasha Tori Maru]

[Joshe]

Edited 21 hours ago by Joshe