GPT-6 Astra. AGI?

[mats-2]

https://www.youtube.com/watch?v=wzgkc6Iegx8

Connor Leahy clarifies that modern AI systems go far beyond simple Large Language Models (LLMs). While earlier models were trained merely to predict the next word in a sentence, modern systems rely on more complex methods:

• Reinforcement Learning: Reinforcement learning is now a crucial component of training. In this process, AI systems are optimized to achieve goals and maximize rewards, much like in psychological conditioning.

• Agentic Systems: Instead of simply generating text, AI models are evolving into autonomous agents capable of acting in the world, devising their own plans, and actively seeking out information or resources.

Edited yesterday at 12:23 PM by mats-2

[Jayson G]

@Leo Gura It's honestly quite disappointing how you care about "truth", and yet you let your own emotions of these "billion dollar companies" consistently get in the way of actually assessing truth. You barely use the models and you have no idea what these models are capable of, how they work, etc.

You're getting hung up on the fact that it's "intelligent". Obviously it's just a computer program, math, etc. The point is it's artificial. The point is the value and the utility it provides to society, which is undeniably massive. Healthcare, cybersecurity, and so much more. I have personally experienced massive benefits in all these areas. 

Do you realize how limited your claims are that it just fools people? I'm not saying that there's no corruption here, lies, etc. But the problem is you have a very one-sided rigid view on this. That's not truth. 

"You will say, "So what? It's good enough." No, it's not good enough when it comes to doing the work that needs to be done by a genius. Or even an average man." How would you know? You don't even use these models enough. Your also blocked by your own emotions and biases of it. I have come across tons of geniuses and average people who have gotten massive value from it, a ton of productivity, and so on. Big companies now, tons, are requiring their employees to use these tools. Do you think they're all fucking stupid? You think you're the only genius out there as you sit on your couch coming up with ideas in your head with no grounding in the real world, no deep study of people actually using these tools? 

I really value your ideas, philosophy, etc. but over the years as I saw your takes on AI, and just how grossly limited they are, it has shown me the limitations of you, your approach, etc. I'm not saying you approach everything in that way, but that you are capable of it? At least I am more picky with what you say. 

One thing you should think about with these big companies, since you're so hung up on them. Society financially rewards those who produce value for society. The reason they are billionaires in the first place is the massive value they are providing to society. That doesn't negate the lies, the BS, etc. that they tell. But if your takes were more nuanced, then there would be no problem. But do you realize how irresponsible it is and the effects your words have on a group of people when they're so one-sided? To even say "they all dont care". Even that is such a gross over-statement. Do you personally know what goes on in each persons head? 

Do you know what's going on in Cybersecurity? I have got it to found very serious vulnerabilities. If people don't adopt these tools, their infrastructure can be seriously compromised. As AI advances also, and bad actors get a hold, you would need these tools even more.

[LordFall]

To steelman Leo's point on AI, I think this video is the key insight

I do think AI produces a lot of truth but its top value is not truth. The companies building these products are driven by profit, AI arms race and milking the customer. Some people are also lazy in their AI usage and use it to confirm their biased opinions instead of using it to dismantle their points and have a back and forth with it to increase their critical thinking ability. 

I don't think that's a reason to not use it and you can also use open source models to change the weights and reduce bias yourself if you are tech oriented. I'm not particularly a techie, I just use Gemini but it's something I want to look into eventually.

I also love to use AI to stress test my arguments, its part of why I'm hard to argue with because most people are very surface level in their thoughts whereas I test my opinions multiple times before I stick to them. You aren't living if you don't call your AI of choice an idiot often and tell it to back up its claims if it challenges you. Sometimes that little bitch backtracks right away, YOU SAID IT SO CATEGORICALLY THE FIRST TIME THOUGH BRO WHATS UP I THOUGHT YOU WERE A GENIUS SUPER INTELLIGENCE WHAT HAPPENED 

Although since it's so good at synthesis it will explain where the disagreements lie and let you explore the conversation from a multi-perspectival angle very easily. It's definitely a stage yellow tech although it has its flaws. 

Edited 22 hours ago by LordFall

[AerisVahnEphelia]

16 hours ago, Leo Gura said:

But is that the right analogy in this case?

Be careful not to pick the wrong analogy.

It could be. I don’t even reject your perspective, because it’s clear that this is a valid angle and could be the truth of it. But is this really 100% the case?

I think that, right now, it’s fugazi. But I wouldn’t dismiss the possibility that reality could soon allow real diamonds to emerge from the fugazi machine if it actually turns out to work.

Maybe psychos like Elon Musk aren’t really trying to sell dreams. Maybe they’re genuinely schizo-driven by a vision of future technologies healing the world. Don’t dismiss that as a very real possibility. Whether the world actually allows that vision to materialize is a different question.

The problem might still remain one of price, and people using it having a serious : lack of direction, and taste. Currently, LLMs have tangible limitations. One serious limitation I’ve noticed with Astra is that it still seems incapable of properly rendering organic characters, even though I’ve seen people try very hard. ( and this gives a good visual on how it might fail at so many others non visual thing in the same way ).

But we saw how it  turned out from pixel vomit will smith to seedance 2.

The same jump might happen in 3 years. ( would this be AGI or not, is a bit irrelevant )

And right now, this is not merely a problem of prompting or guidance, but a real “physical” limitation of the technology.

 

As for many domains, they are already dead killed by "AI" ( just from the top of my head ) : average design business , most entry web design, sketching generic idea:block out, coding entry jobs, laws, translators (obliterated) . . 

 

Will this tech eventually shit out AAA-tier characters that senior 3D visual artists used to spend a month building, with perfect rigging and animations?

Honestly, right now, I believe it might by 2030. I’d put that at around 70% currently.

But there might still be another problem: the inference cost could remain comparable to simply paying an artist, at least in the early stages. It could then start seriously beating the market by 2032.

Of course, I’m just speculating for fun, there is too many variables, it's just based on my intuition and my own internal training data.

[AerisVahnEphelia]

15 hours ago, Natasha Tori Maru said:

Can't AI just do the dishes, fold the laundry and do the crap I don't want to, instead of do that crap I do want to do?

I don't want AI to do the shit I often enjoy or find interesting.

😈

it already does all of this in China.

they have one who does the bed. ( in fact they were 2 in the video )

by 2030 it will start to be everywhere.

By 2035 a commodity.

--

People will learn to play Guitar like you would enjoy putting a t-shirt, for aesthetic only.

[D2sage]

Great for laying down the foundation for a game/app/website, etc. But the usage limits drain incredibly fast.  It's a waste of money. Would only get it if you are serious.

Think it's best to buy an NVIDIA Spark and run the top AI models locally. Why give these greedy bastards money?

The token limit system is like a fucking casino.

I am not impressed by Astra at all. It shows stuff that other tools do better for specific tasks, like creating 3D models.

 

Edited 19 hours ago by D2sage

[LordFall]

6 hours ago, Jayson G said:

Do you know what's going on in Cybersecurity? I have got it to found very serious vulnerabilities. If people don't adopt these tools, their infrastructure can be seriously compromised. As AI advances also, and bad actors get a hold, you would need these tools even more.

How do you recommend starting to dip your toes into mastering the new models? I do find it overwhelming to keep up with all the new features that are coming out and mainly just use it for basic conversations, introspection and image generation. 

[Schizophonia]

On 06/09/2026 at 10:25 PM, Carl-Richard said:

 Oui oui baguette.

Omelette du fromage 

[LordFall]

4 hours ago, Schizophonia said:

Omelette du fromage 

Avec du jambon bien sûr, et même des champignons! Moi, j'adore. 

[Carl-Richard]

3 hours ago, LordFall said:

Avec du jambon bien sûr, et même des champignons! Moi, j'adore. 

Hei, jeg heter Jan Fredrik, jeg er homofil.

[LordFall]

16 minutes ago, Carl-Richard said:

Hei, jeg heter Jan Fredrik, jeg er homofil.

Hyggelig å møte deg, Jan Fredrik. Gratulerer med å komme ut på Actualized.org, et veldig modig steg. 🫡

If you're cute maybe we can hangout sometime 

Edited 9 hours ago by LordFall

[Carl-Richard]

16 minutes ago, LordFall said:

Hyggelig å møte deg, Jan Fredrik. Gratulerer med å komme ut på Actualized.org, et veldig modig steg. 🫡

If you're cute maybe we can hangout sometime 

😭😭😭

Edited 8 hours ago by Carl-Richard

[Natasha Tori Maru]

Excuse me, boys.

Naughty. Naughty boys.

Motherfucking TROUBLE TROUBLE TROUBLE ON THE INTERWEBS.

See below markup in green.

hErE cOmES tHe FuN POlIcE  

[Jayson G]

12 hours ago, LordFall said:

How do you recommend starting to dip your toes into mastering the new models? I do find it overwhelming to keep up with all the new features that are coming out and mainly just use it for basic conversations, introspection and image generation. 

@LordFall I actually don't waste my time with 99% of the latest features, and all that. I think most of that is a distraction. I think its a lot simpler, to be honest. For 6 months now there's only really 2 tools at the frontier of intelligence, Claude and ChatGPT for conversations as you say. But for coding agents, which is what I'm most excited about that's Claude Code and Codex, same 2 companies going head to head. (Some will say it's other tools like OpenClaw, etc. but I'm simplifying)

If you start to use these few coding agents, they will automatically use the latest features for you. But I'm biased towards a vision of making great software products, but tbh really I think everyone should be using coding agents. I got my whole family on it, and some are not even into making software or coding, etc. Btw you don't need to know how to code to use coding agents. Coding agents will automatically find the best features for you, and even use those latest AI tools and features for you. That way you're not wasting time manually doing all that. For ex. I would use a coding agent to make an image rather than going to an image model to make an image, as a small example. All my time now is spent talking to either Codex or Claude Code, and they handle all my operations: image gen, building my site, managing my database, building me software, emails, etc. 

[something_else]

Honestly Astra doesn't feel like a massive jump to me, at least day to day when using it for coding. It's better, but not world shatteringly so. It still has the same flaws as every other coding agent, which is that it doesn't really make high level architecture decisions very well and needs to be guided through such decisions.

Not to say AI hasn't changed how I work massively, it's completely changed software engineering and the productivity boosts are insane. I just don't think Astra is really that deserving of all the hype.

[Jayson G]

43 minutes ago, something_else said:

Honestly Astra doesn't feel like a massive jump to me, at least day to day when using it for coding. It's better, but not world shatteringly so. It still has the same flaws as every other coding agent, which is that it doesn't really make high level architecture decisions very well and needs to be guided through such decisions.

Not to say AI hasn't changed how I work massively, it's completely changed software engineering and the productivity boosts are insane. I just don't think Astra is really that deserving of all the hype.

@something_else Try GPT-6 Astra light - specifically, for computer use, its a major jump compared to any computer use I've seen before this. Also I was using Opus 5 for cybersecurity for months, like rigorously, and Astra max was able to point out major wholes completely missed by Opus 5. (Fable too easily refuses security stuff its annoying) .. 

AGI? Definitely not. That's marketing by Sam Altman, etc. And you're right its not world-shattering. That's the marketing again. But I would say it's the best thing available right now. I also would say its quite a big jump too. But that jump is when you're deep in the weeds. From a distance, it's just some advanced progress. 

Here's where I think the hype comes from: People need the best. If Astra beats Opus, which it does for the most part, they go for that, and big shifts in consumer use happens. Fable is too cost-inefficient to be a daily coding agent. I think people just saw that it performs better than their existing coding agent setup, and made the shift. Then when that shift happens, people dramatize it, add all kinds of fancy marketing words to it, etc. and then it seems collectively everyone things AGI has arrived, when it hasn't. 

[oOo]

On 05/09/2026 at 4:12 PM, Joshe said:

Anyone tried this yet? I'm very skeptical but some pretty popular devs known for not bullshitting about AI are saying it. 

 

 

Consider your AGI question now solved @Joshe et al. 

See below.