Claude Mythos is crazy good

[AION]

And we are just in the beginning of the AI boom. Just imagine AI in the wrong hands. 

 

Edited April 8 by AION

[cistanche_enjoyer]

Yes it's crazy. We are in full acceleration.

[NewKidOnTheBlock]

The names themselves keep getting increasingly divine as well. What's gonna be the next model's name, Yahweh?

[Xonas Pitfall]

9 minutes ago, NewKidOnTheBlock said:

The names themselves keep getting increasingly divine as well. What's gonna be the next model's name, Yahweh?

🤣🤣🤣

[bazera]

 

 

[Hayato]

It is possible that there are much more advanced AI models that are kept hidden and not shown to the public.

[Natasha Tori Maru]

47 minutes ago, Hayato said:

It is possible that there are much more advanced AI models that are kept hidden and not shown to the public.

At the risk of sounding conspiratorial - I don't see why the military, for example, wouldn't have some sort of early access. Domain specific models more likely.

Military often have early access to new and cutting edge tech.

[AION]

1 hour ago, Hayato said:

It is possible that there are much more advanced AI models that are kept hidden and not shown to the public.

Yes, it is called Cluade Mythos

[something_else]

Pretty scary. Seems Anthropic are going about it the right way though.

[AION]

Opus model is already very good. Night and day compared to other mainstream AI. 

[MutedMiles]

3 hours ago, Natasha Tori Maru said:

At the risk of sounding conspiratorial - I don't see why the military, for example, wouldn't have some sort of early access. Domain specific models more likely.

Military often have early access to new and cutting edge tech.

Yes, the models are increasingly becoming domain specific after Claude success at coding, that's basically all the recent success of AI is about, domain specific skills and tools.

Edited Friday at 12:56 PM by MutedMiles

[something_else]

40 minutes ago, AION said:

Opus model is already very good. Night and day compared to other mainstream AI. 

Yea, it's pretty insane how good it is at coding. It does still make incredibly stupid decisions sometimes and definitely still needs human direction and oversight. But overall it has completely changed how I work as a software engineer.

[bazera]

5 hours ago, something_else said:

But overall it has completely changed how I work as a software engineer.

Yeah, software industry is being reshapend in front of our eyes. Hard to say where all these will take us.

[OBEler]

This is the biggest step. We have reached reasoning. Something experts never thought this would be possible with LLMs without mayor change in architecture.

What are the consequences. 

 

[Breathe]

On 4/10/2026 at 1:24 PM, something_else said:

Yea, it's pretty insane how good it is at coding. It does still make incredibly stupid decisions sometimes and definitely still needs human direction and oversight. But overall it has completely changed how I work as a software engineer.

I took a one year break from coding and just got back in the game. I was very impressed with Claude but was simultaneously perplexed by it's ability to make newbie mistakes. It go from master level developer to junior level developer in an instant. 

For the people out there saying that this is going to blow over, it's not. This shit is just way too intellgent and capable.

When I first started learning about AI I thought, man, it's going to be the best hacker that's ever lived. Security and exploitation is one hell of an application for AI.

Edited 4 hours ago by Breathe

[integral]

Im not buying the hype. Ill believe it when i see it.

The hype is ridiculous with Ai, they are always bullshiting. 

Edited 1 hour ago by integral

[UnbornTao]

23 minutes ago, integral said:

Im not buying the hype. Ill believe it when i see it.

The hype is ridiculous with Ai, they are always bullshiting. 

Fair.