VentureBeat

In case you guys are not following him, you need to be because he writes great cybersecurity articles.

He is a senior cybersecurity contributor, in fact, for VentureBeat.

And beyond that, in his free time, he's an adjunct professor at Webster Loyola Marymount University.

So I appreciate you being here.

They're also, they're delegated, they're over-delegated privileges.

And then finally they're extremely autonomous.

And I did a panel at VentureBeat Transform earlier this month and the CISO of of Box was there, and she was saying that that ephemeral nature and that delegated nature of privileges was also brought up by a GM from Rubrik.

And together, their two comments really, I think, paint a portrait of agents in transition and you know, these are two very advanced companies, Box and Rubrik, and they had been trusting the agents and the CISO of Box put it really well.

You're listening to It's About Data, episode 93.

Our guest this time is Rob Strecke of VentureBeat.

And this is actually a really very timely conversation we're about to have with my old friends in the industry, Rob Strecke, who's someone I've seen around basically.

Now you're with VentureBeat and developing the analyst and media business there.

But I've also seen you go to other channels in the past.

More to the point, you are basically a very frequent, I mean, if I'm going to a conference, if I don't see you there, there is definitely something wrong.

I built a service at AWS.

I look at this as I just love helping companies and organizations break down what is so confusing about what everything AI and data and what it all means to them.

This just means the computational graph, the amount of transformer kind of like reasoning thing that's going on in there, which means that they are not doing like multiple loops over this to prevent folks from understanding.

Folks, we have like a leaked VentureBeat article that looks like there's no CSS in here because the internet is down, but maybe...

Maybe their CMS systems were also down, so they weren't able to push it back behind the NDA.

But here's what they say.

Welp, say hello to Claude Fable 5.1 and Mythos 5.1.

Quoting VentureBeat, the two names refer to the same underlying model.

Fable 5.1 is the generally available version with Anthropix production safeguards in place.

Mythos 5.1 is available through restricted access programs for vetted cybersecurity and life sciences organizations that need capabilities normally constrained by those safeguards.

The short in it is ETB.

It was published in SF Weekly, Venture Beat.

I'm forgetting other media.

Big media, US media, Wired, I think, and GeekWire.

And before anything ships to production, a human still signs off.

Now, Cognition, the company behind Devon, they walked through a live demo of this with VentureBeat.

Now, in the demo, when anybody reports a broken feature in the engineering channel, Devon reacts with an emoji, replies right in the thread, investigates, and opens a pull request.

It even knows who the code owner is, so it tags the right person.

Anyway, maybe from that comment, you could tell Stegman is a product person.

She absolutely is.

And she talked about her own experiences using Slack code in that VentureBeat briefing.

When she puts up a pull request, she tags in an engineer from her team before asking for approval.

We're going to dive into this in a lot more detail.

But speaking of AI, and before I get to the guest, one other thing that's really interesting, this is an article that just recently was published in VB VentureBeat Weekly.

So VentureBeat.com is the website.

And this was just published a few days ago.

And the headline is, Enterprises with AI Context Layers Report Agent Failures at More Than Twice the Rate.

Perplexity has launched Portable Computer, a local AI agent platform running fully on-device with zero token costs, starting with NVIDIA DGX, Spark, and RTX Linux PCs.

Quoting VentureBeat.

"The launch, developed in close partnership with NVIDIA, is one of the most aggressive attempts yet to move serious AI agent workloads off the cloud and onto local devices.

The model, the user's files, and the work itself can all stay on the machine.