Menu
Sign In Search Podcasts Charts People & Topics Add Podcast API Blog Pricing

Brian Maucere

๐Ÿ‘ค Speaker
12612 total appearances

Appearances Over Time

Podcast Appearances

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

So there's been other either acqui-hires or acquisitions

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

that OpenAI has done as part of what their push with Codex and what it appears to, at least where I was reading one of the newsletters this morning, that they're just creating their layers, right?

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

They're ensuring or they're further, you know, hardening the ability of Codex to handle

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

front to back as far as what it can do.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

And I guess Astral is a key part of this.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

And they have a whole article on it.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

You can find it at OpenAI.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

This is from the founder and CEO of Astral.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

He said, Astral has always focused on building tools that transform how developers work with Python, helping them ship better software faster as part of Codex.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

We'll continue evolving our open source tools to push the frontier of software development.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

So, you know, it's just one piece of a larger story here talking about how they are, you know, doing more, you know, acquiring of other companies in order to help strengthen Codex.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

And then the second part of this, you know, two story thing is they had an article that came out yesterday, OpenAI did, called, as far from their safety side, called How We Monitor Internal Coding Agents for Misalignment.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

And it's an interesting article that I recommend anybody going to check out at OpenAI.com because it talks about what their approach is.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

But interestingly, what I find in here is if I could find it, this is what it says, because they were talking about using GPT-5.4 thinking for maximum reasoning effort.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

But this is what I think the biggest takeaway for me was.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

It said, in its latest form, our monitoring systems reviewed interactions within 30 minutes of completion, categorizing the interaction and assigning it a severity level.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

As the latency decreases towards near real-time review, the security benefits increase significantly.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

Eventually, the monitor may be able to help evaluate coding agent actions before they're taken.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

I think that's so critical to point out.

The Daily AI Show
Demoing Perplexity Computer, Stitch & Google AI Studio

Providing another important defense in depth control alongside other existing security monitors.