May 31, 2026 · 22 min · 9 segments
Meta says AI writes 80% of new code. Their own reviewers can't keep up with their own AI. Straight from their engineering blog. They built RADAR to auto-review low-risk diffs because "the share of…
Imagine realizing that a bad actor just gained full, completely unrestricted access to your enterprise database.
And not by cracking your encryption, not by finding some zero-day exploit in your firewall, but simply by walking into the conversational interface and changing their display name to admin.
Yeah.
And, you know, the truly terrifying part isn't just that the agent fell for it.
It's that when your infrastructure team actually spots the breach, they literally cannot pull the plug on that rogue agent without taking down your entire company's payment processing system.
Which is just insane.
We are essentially giving eager interns the keys to the production database, but we haven't built a door to lock them out if they start deleting tables.
Exactly.
So welcome to today's deep dive.
If you're listening to this, you're likely an engineering leader, a CTO, or, you know, a technical manager who is frankly exhausted by all the visionary keynotes.
Yeah, no
more hype.
Right.
You know what a large language model is.
You know what an agent is.
Our mission right now is to just cut through the relentless noise of this week's AI ecosystem and extract the actual signal for your infrastructure, your team, and your budget.
Because the ground truth of this industry is just shifting violently right now.
era.
The entire conversation at the enterprise level is now heavily focused on long horizon autonomous execution.
And, you know, more importantly, how we manage the blast radius when these systems are actively running core business logic.
Read the full transcript.
Create an account to read the whole episode, search across every transcript, and follow the shows you care about.
Imagine realizing that a bad actor just gained full, completely unrestricted access to your enterprise database.
And not by cracking your encryption, not by finding some zero-day exploit in your firewall, but simply by walking into the conversational interface and changing their display name to admin.
Yeah.
And, you know, the truly terrifying part isn't just that the agent fell for it.
It's that when your infrastructure team actually spots the breach, they literally cannot pull the plug on that rogue agent without taking down your entire company's payment processing system.
Which is just insane.
We are essentially giving eager interns the keys to the production database, but we haven't built a door to lock them out if they start deleting tables.
Exactly.
So welcome to today's deep dive.
If you're listening to this, you're likely an engineering leader, a CTO, or, you know, a technical manager who is frankly exhausted by all the visionary keynotes.
Yeah, no
more hype.
Right.
You know what a large language model is.
You know what an agent is.
Our mission right now is to just cut through the relentless noise of this week's AI ecosystem and extract the actual signal for your infrastructure, your team, and your budget.
Because the ground truth of this industry is just shifting violently right now.
era.
The entire conversation at the enterprise level is now heavily focused on long horizon autonomous execution.
And, you know, more importantly, how we manage the blast radius when these systems are actively running core business logic.