Brian Maucere

Spark, the current level of Spark is not open source.

They might release a smaller version that they will open source, but they're not entirely sure about that.

And they did that.

They sort of announced that around the Llama series, too, like, hey, we we did this, but we're not committing to doing it moving forward.

Right.

Also, that was very much Yann LeCun's vision, right?

I'll come and be your AI person and give you legitimacy, but these are my standards and that the last llama didn't really work and somebody else was hired, Yann left.

But the other thing, Carl, is interesting that they chose llama to build.

Did they intentionally not choose the Chinese models or was it just not available in terms of timing?

So I found out because Jumi was on an after show, long heated conversation about, you know, limiting how he feels about anthropic, reducing the limits and and how he hits them within a couple hours.

And his big issue was the weekly limit.

What what perplexity was doing was the.

monthly limit, I think, because what the rant was, the Reddit rant that was talked about was basically from the perspective of a researcher, an academic researcher.

And if we remember, Perplexity's roots were about getting academics really excited because it was a search engine that searched papers and you could use the sources as

actual accurate sources was the only AI that was doing it at the time.

It used to be that with a pro subscription, which is the plus level, right?

The $20 subscription, you could do hundreds of more than a hundred, sometimes up to 300.

It seemed deep research queries.

Now you can do 20 within a month.

And that's a steep drop, right?