Open-source intelligence

That's called Architect.

They offer guardrail-free conspiracy discussion, uncensored image generation, and OSINT you can only purchase with crypto, but they claim this is only for research purposes and shouldn't be used to do anything illegal.

So how does this work? Well, the researchers think it uses Grok for its primary inference, uses Google Cloud and Cloudflare for hosting, It uses Claude for long context model layer, Llama routed through open router for certain specialized tasks, Tavoli for live search, now payments for crypto, Cloudflare and Let's Encrypt for DNS and TLS, and this distributed architecture makes it very hard to take down technically, and their claims have made it very hard to take down legally.

And also because it's so distributed, it's really hard to assemble a case.

mean, is that normal protocol or was the system just completely broken from the start?

Well, that is the multimillion euro question, isn't it? In modern intelligence, agencies constantly monitor open source intelligence or OSINT.

Right.

OSINT.

Exactly.

Which includes public telegram channels and WhatsApp groups.

Okay, now, th- what he showed you here in this, this is from Tendar.

Tendar is a, um, uh, an OSINT analyst.

The Russian salient north of Liman, the- the Donetsk region, uh, Ukraine has collapsed.

Ukrainian forces are moving into this area, conducting mopping-up operations.

mean, we'll see on troop movements and that stuff will drop because none of this is secret.

It's super easy for the surrounding areas to see troop movements and put that stuff on OSINT.

So we'll see for sure.

And you can't hide a big troop movement today.

They know how the midterms are going to go.

Our enemy is not stupid.

You have OSINT too.

They can figure this stuff out and they're just waiting.

Right.

So if you get too wedded to using open source intelligence and you might convince yourself that what I'm getting through OSINT is the truth.

And then maybe what I'm getting from my military systems might be false because it's been deceived by jammers or electronic warfare.

So you can use that OSINT as another avenue to introduce false information into the commander's decision-making process.

So it can be sort of a two-edged sword where it's very useful now, but it could become a very unuseful source in the future.

For a slightly different application of AI in a military context, Anthropic released this long stream of all the ways that suspect actors around the world have been, you know, playing Claude instead of making, you know, organizing apps, apparently.

Which is where the whole thing is heading.

The same way Daniel's automation work is about removing the human from the loop, the OSINT community is automating the detection of sanctions evasion.

the ships can't hide from a model that watches every pixel of every pass

the ships aren't getting smaller a very large crude carrier is the size of a small city block it can't hide it can only lie and the lie is getting easier to catch

I had someone look at his stock two years ago and said, damn, I could do that work.

You know why? It was kind of like related to the kind of social media OSINT stocking that everyone actually does.

Let me check this photograph.

Who was in this photo? Let me zoom in here.