Sweet Attack: How Sweet Security Proves Real Attack Paths With Runtime Intelligence

Chris Lentricchia, Product Marketing at Sweet Security, returns to the show to break down Sweet Attack, the AI red-team agent that doesn’t model the attacker, it is the attacker.Running continuously on live runtime data, Sweet Attack chains together real vulnerabilities to prove exactly how an attacker could breach your environment, then hands over evidence and steps to reproduce.Chris also walks us through the attack-fix-defend learning loop and why proactive, automated security is the only answer to AI-speed attackers.If your team is drowning in theoretical vulnerabilities, this one’s for you.

David Wang, Head of Product, Marketing, and Customer Success at Tetrate, joins to break down how enterprises can actually manage AI at scale.We unpack the anatomy of an AI request, how organizations can track spend and prove AI ROI, and why Envoy-based AI gateways are the key to guardrails, policy control, and developer agility without adding latency.If your team is trying to rein in model costs while keeping teams moving fast, this one’s for you.

Eyal Bukchin, Co-founder and CTO at MetalBear, makes the case that AI agents write code fast, but they build on assumptions about your stack that may be wrong.He shares how mirrord gives agents real cloud context with preview environments, DB branching, and chaos-style failure simulation, and makes a bold prediction about Kubernetes’ future.If your team is adopting Cursor, Copilot, or Claude and wondering how to trust what agents ship, this one’s for you.

Lawrence Michael and Stephen Dendtler, co-founders, CEO and CTO of RackCorp, join the show to break down what sovereignty really means for cloud infrastructure: data that stays in-country, operations run under one jurisdiction, and no kill switch that lets a foreign vendor or government cut off a country’s critical services.They walk through OneStack, the fully bootstrapped platform powering 35 data centers across 16 countries, why banks and governments are ditching VMware, and how hosted LLMs are bringing sovereign AI to regulated markets.If your organization has compliance or jurisdiction requirements that hyperscalers can’t meet, this one’s for you.

al Ordo, co-founder and CPO at Native Security, argues that the era of detection-based cloud security is over. In a world where AI models like Anthropic’s Mythos can find and exploit vulnerabilities in minutes, the only viable defense is secure-by-design architecture, what Native calls “shift down” security.If your cloud team is still relying on alerts and hoping to respond in time, this one’s for you.

In this episode of Software Plaza, we're joined by Emanuela Zaccone, AI and Cybersecurity Product Strategist at Sysdig, and Eric Carter, Director of Product Marketing at Sysdig, for part two of our deep dive into the future of agentic security. They walk us through Sysdig's new Headless Cloud Security product — an agentic approach that lets security teams use AI coding agents like Claude Code and Codex to investigate vulnerabilities, triage runtime threats, and automate remediation, all without touching the traditional UI. The results speak for themselves: what used to take three people 45 minutes to 2 hours (at $400 per incident) now costs $3.75 in tokens and takes minutes. If your team is drowning in security alerts and looking for a smarter, faster way to respond, this one's for you.

Rohit Agrawal, Senior Staff Software Engineer at Databricks, shares how the traffic platform team scaled Envoy from 1,000 to billions of QPS, why they chose sidecarless proxying over Istio, and how dynamic modules are set to double throughput by eliminating sidecars entirely.If your team is wrestling with service mesh decisions at scale, this conversation is packed with hard-won lessons.

Mo’ath Qasim, VP of Engineering at Kubermatic, explains why enterprises are ditching VMware after Broadcom’s acquisition and migrating to KubeVirt-based virtualization on Kubernetes. He walks through the Kubermatic Virtualization dashboard, migration tooling, and how the platform delivers a familiar VMware experience with the power of Kubernetes underneath. If your team is rethinking your virtualization strategy after Broadcom’s price hikes, this one’s for you.

In this episode, Amod Gupta explains why AI-generated code is functionally correct but often insecure, why traditional scanners fall short in the age of agentic coding, and how Endor’s AURI platform provides a security intelligence layer that works alongside coding agents like Cursor and Claude Code.He walks through real-world incidents including dropped databases and production wipeouts, and shows how purpose-built tools can shift security left to the point of code generation.If your team is using AI to write production code, this one is for you.

Lauren Morris, Head of Product at Amazee, shares her journey from library science and improv to building AI infrastructure that puts data sovereignty first.She walks through Amazee’s AI Gateway — a multi-region, enterprise-grade API gateway for LLMs that keeps data within geographic boundaries — along with built-in cost controls, team-level governance, and observability across 30+ models from OpenAI to open-weight.If your team is navigating the tension between AI adoption and data compliance, this one’s for you.Amazee has provided a coupon code AMAZIC10 for $10 in free tokens for our SoftwarePlaza viewers.Just visit Amazee.ai, select the AI Gateway product and paste the coupon code in the “Add promotion code” box in the checkout cart.