00:00I'm here with Amon, co-founder and CTO of HashiCorp.
00:03How has reInvert been for you so far?
00:05It's been fun.
00:06It's been exciting.
00:06Lots of good energy, fun announcements.
00:08So based on the recent announcements, we've heard a lot about agenting AI.
00:12How has HashiCorp been using this to help customers?
00:15We're looking at everything from bringing MCP agents for tools like Terraform and Vault
00:18and Packer and Console, but probably 30 to 40% of Terraform code is already being written
00:22by agents or coding assistant tools.
00:24So I think the adoption of some of this stuff has been super, super fast.
00:27For us, we're looking at, okay, how do you start applying that, not just for day one
00:31of the lifecycle, right?
00:33You can say, great, I have a blank app and I want to set up my initial cloud environment.
00:36You can use tools to write the Terraform for you.
00:38But it's also looking at the day two, day three operations of existing managed infrastructure.
00:42So that's been an area with things like Project InfraGraph, where we're really looking at how
00:45do you bring the context of your existing infrastructure so then you can actually start applying those
00:50kind of AI ops techniques across the full lifecycle.
00:53Amazing.
00:53So one question that many customers have when they adopt Agentic AI is how can we maintain
00:58the security without running into any issues?
01:01How does TashiCorp address this?
01:02I think for us, our view is it really comes back to the notion of policy as code.
01:05So how do we create automated guardrails to say, okay, if you're having an agent write,
01:09let's say, the Terraform code, you're not going to have a human then sit there and review
01:12the 10,000 lines.
01:13So instead, you have to codify your policies.
01:15And those could be security policies, operational policies, compliance policies.
01:19So this has been a big effort we've done with Amazon.
01:21We announced a strategic collaboration agreement with them where we've been jointly developing
01:25a set of these policy libraries for common frameworks.
01:27So the goal becomes how do you create sort of a safe sandbox where the AI can kind of play
01:31within that, but if it tries to sort of break the sandbox or violate any of those policies,
01:35we can stop it in a programmatic way.
01:37I love the idea of a sandbox.
01:38So in the age of AI, what's one skill that we can all build right now to stay ahead?
01:43I think what sort of never goes out of style is understanding the fundamentals.
01:46You know, if you're building a modern, you know, let's say JavaScript application,
01:49if you just tell the AI, hey, build me a JavaScript application that does something,
01:53it's sort of underspecified.
01:54You're leaving too many degrees of freedom versus if you knew the technology at a slightly
01:58better level, you might say, hey, I want to use the React framework and I want to architect
02:01it as a model view controller.
02:03And here's how I want the app to be.
02:05That higher level of specificity allows it to actually generate a much higher quality output.
02:10And so in that sense, I think the skill to learn is still the same one.
02:13Having a deep understanding of the fundamentals of core architectures,
02:16core application building platforms, you know, scaling patterns and behaviors allows you to
02:20be much more specific in the, in the guidance you're giving AI and much more productive.
02:24Therefore,
Comments
2