Leo's Blog Discussion Mega-Thread

[Natasha Tori Maru]

@Leo Gura My god - this portly fucker is gonna break skippies legs  

 

[Jayson G]

@Leo Gura To even claim they're over-invested means you understand what they're working on, the convergence of these technologies, that you're deep in the weeds of seeing what's working and what's not, but can you honestly say you're doing all that? To speak from a very high level philosophical view is, in my opinion, limited and biased. 

To say they're enslaving, well you might be right there. I already have friends enslaved by these companies lol 

But if they weren't enslaved by these companies, they'd have a different kind of shittier life. 

[Leo Gura]

25 minutes ago, Jayson G said:

@Leo Gura I don't know if you realize this but you go back and forth a lot between "we will have insane robots", "its quite possible there will be massive job loss", etc. and that all this is just hype and stuff. 

1) It depends a lot on timeline. In 10+ years I can see a huge robotics revolution. But these companies are already shipping robots to consumers today. 10 years is a large gap.

2) I have changed my mind about job loss. I no longer believe that will happen.

Quote

But you're right, we don't know if this is going to work. But if you look at actual evidence that Figure put out, its actually doing a) real work at BMW, and b) continuous progress, like in its folding laundry example. It's showing learning behavior in things its not taught. 

The problem is that none of these tech demos can be trusted to be genuine. These companies have zero qualms lying to you. They will have a slave operate the robot and call it autonomous. You can never trust they aren't doing that.

Musk is clearly doing that.

Quote

Often you categorize all these people as just "tech bros", and their is understandably a lot of corruption and hype, but you're underestimating what they're doing, what they know, etc. Do you think they're stupid to dump so much money in all this? They did before things panned out. Now if you look at Gemini for ex. its genuinely phenomenal at coding. People are making video games on it, when you said it's just good at websites for example. It's constantly proving people wrong. 

You are right that they are smart in certain ways, so we must be careful not to get too cocky or we will look like the fools. However, they are also shameless liars, as Musk has proven. And their companies are crazily overinvested, making them desperate for money.

No, LLMs are not great at coding. They make so many mistakes that serious coding with them is impossible. It's a facade. Dig deep and you will see so many problems.

Vibe coding is a lie.

But hey, it's not a bad idea to be skeptical of my views on AI. I could be wrong. Predicting the future is a bitch.

Edited 15 hours ago by Leo Gura

[Jayson G]

@Leo Gura Your claim that they are bad at coding is so limited. I code with them on a daily basis. It's gotten so good to the point that I can make a fully coded app front-end without a single error, in 1 shot prompting, and the design is unlike anything I see online. This was possible in the past 1 month with Gemini 3 pro release. You are talking about tech that's 8 months ago. 

When was the last time you actually tried coding with it, because if you didn't your opinion is out-dated and philosophical. 

And sure, they lie a lot, but I'm looking often at what I can see with my eyes - the laundry example in robots, the video game that Pieter Levels made with AI, etc. 

Edited 15 hours ago by Jayson G

[Leo Gura]

@Jayson G 

 

[Jayson G]

@Leo Gura your evidence is a video of a guy who says the stuff that I experienced myself as well many months ago. I get where he's coming from. But again, just last month things changed a lot. 

https://geo-void-explorer.lovable.app 

Try this for example. It's a 3D space, with beautiful visuals, etc. I did this with a single prompt. It's fluid motion, and no errors. I literally just did a single prompt with Gemini 3 pro. Before this it was doing terrible design, buggy code on basic CRUD apps, etc. 

These opinions are a bi-product of the tech of where its at right now, but to react at what's there right now and think that that's the future is premature. 

And keep in mind, I'm not a coder and stuff. For people deeply involved, they are getting 100x more value than me. People who are frustrated just don't know how to use the tools, and are also using wrong tools, or expecting something when the tech isn't ready. They are actually improving and I constantly see that over time. To think also that it's used up all the data of the internet for training is also wrong. Clearly it used that up a long time ago so then why are many areas of application dramatically improving?

Edited 15 hours ago by Jayson G

[LoneWonderer]

35 minutes ago, Natasha Tori Maru said:

I do not have a car.

Do you live in or near a city in AU? I'm just wondering if AU is anything like NZ because here even if you live in Auckland, public transport is bad at best and very basic public transport rest of country, makes it hard to move around.

Back in Europe I would NEVER own a car but here in NZ it's quite difficult if not impossible.

[Leo Gura]

13 minutes ago, Jayson G said:

Try this for example. It's a 3D space, with beautiful visuals, etc. I did this with a single prompt.

Coding basic good-looking websites is very simple these days. There are so many tools for this. This does not represent a serious programming project.

AI is not capable of replacing human programmers.

Yes, AI is good at coding short, small, self-contained projects like a little PacMan game or personal website. But it cannot code a professional corporate project. An LLM will never be able to do that.

Edited 15 hours ago by Leo Gura

[Leo Gura]

Look, I could be wrong. Please form your own conclusions.

But I am pushing back on the AGI hype train. Calling their bluff.

Let's see who's right in 3 years.

Edited 14 hours ago by Leo Gura

[Jayson G]

4 minutes ago, Leo Gura said:

Coding basic good-looking websites is very simple these days. There are so many tools for this. This does not represent a serious programming project.

AI is not capable of replacing human programmers.

Yes, AI is good at coding short, small, self-contained projects like a little PacMan game or personal website. But it cannot code a professional corporate project. An LLM will never be able to do that.

@Leo Gura yes right now it can't. But the entire silicon valley right now is using a coding IDE. It doesn't code everything but most serious programmers are using cursor. Y-combinator has evidence on their batches, and they report that more than 50% are using cursor. 

https://fly.pieter.com/ .. this is also a video game made by Pieter Levels. I'm not saying the tech is ultra good enough to do big things. I'm just saying its far better than it was 6 months ago. If it's far better, its more and more doing more capable things. Design has improved immensely. 3D environments as well. 

Corporate wise, AI phone agents are now replacing many actual receptionists, and they are good. AI lead generation systems as well. 

The technology right now is crude yes, but its getting better and better each month.

[Jayson G]

2 minutes ago, Leo Gura said:

Look, I could be wrong. Please form your own conclusions.

But I am pushing back on the AGI hype train. Calling their bluff.

Let's see who's right in 3 years.

@Leo Gura AGI is different from developments in coding. AGI is largely fantasy with some truth and evidence and direction. 

[Leo Gura]

1 minute ago, Jayson G said:

Corporate wise, AI phone agents are now replacing many actual receptionists

 

[Leo Gura]

Just now, Jayson G said:

@Leo Gura AGI is different from developments in coding. AGI is largely fantasy with some truth and evidence and direction. 

I will argue that you can't do serious code without AGI.

[Jayson G]

9 minutes ago, Leo Gura said:

I will argue that you can't do serious code without AGI.

Then why is most of NVIDIA employees using AI-assisted coding? (well thats what Jensen claims at least)

And that phone agent video is quite spot on. I was ordering dominoes many months ago, or when I have to get a visit to the doctor, I always immediately say "Transfer me to a human right now." 

Not saying I like the experience, but I guess it financially does the job for companies, whether its quality experience or not. 

Edited 14 hours ago by Jayson G

[Leo Gura]

13 minutes ago, Jayson G said:

Then why is most of NVIDIA employees using AI-assisted coding? (well thats what Jensen claims at least)

They are paid $3 trillion dollars to lie to you.

You asked the fox in the hen-house if he has eaten any chickens. And you trust him?

Haha

If you gave me $3 trillion dollars, I would use AI all day long. I would have AI clean my toilets. Then I would pay a wage slave on the side to actually clean my toilets while you are gawking at how awesome I am.

Edited 14 hours ago by Leo Gura

[Jayson G]

@Leo Gura I know Jensen is biased actually also. The way he holds trump as god is quite off. But other times he comes off as quite honest. Who knows though? Yeah financial incentive on that level can do that kind of stuff though, so its quite possible. 

I'm quite put off by the AGI phenomena now. I now rarely listen to Sam Altman, Peter Diamandis, etc. Peter and his crew are so so so caught up, as it seems now from my POV, its almost shocking. They get off on the hype train. 

What I'm trying to do is not form an opinion on the actual final conclusion in 5 years because there are traps and lies everywhere. 

Edited 14 hours ago by Jayson G

[Leo Gura]

2 minutes ago, Jayson G said:

he comes off as quite honest.

He is honest. Honestly high on his own supply.

[Jayson G]

4 minutes ago, Leo Gura said:

He is honest. Honestly high on his own supply.

@Leo Gura I don't think we can honestly know what is true and what is not for the most part sitting on the outside sidelines. Hype train bias? for sure. I've seen he's all AI-biased. He has an argument for every possible AI scenario being on the positive side of history. He doesn't think there's a negative scenario in AI for the most part, which sounds too biased.

Edited 14 hours ago by Jayson G

[Leo Gura]

Give me $3 trillion and I will only see rainbows and unicorns ahead.

I'll even lick Trump's asshole as a free bonus!

Edited 14 hours ago by Leo Gura

[Jayson G]

@Leo Gura loll yeah I can see that being possible with him, my biggest internal struggle is sorting truth from falsehood. It's frustrating at times. Reality is messy