Tracy Alloway and Joe Weisenthal host Matt Swisher, founder of OnDB (a data infrastructure startup for agentic AI) and legendary French hacker with 20 years in enterprise cybersecurity. Swisher previously analyzed major leaks including Shadow Brokers and WannaCry attacks.

The conversation explores the intersection of AI and cybersecurity amid escalating tensions between Israel and Iran. Topics include Iran's cyber capabilities, Israel's traffic light hacking operations, and the recent drone attacks on Amazon data centers that caused widespread service disruptions.

Swisher discusses how AI is transforming both offensive and defensive cybersecurity, the emerging threat landscape for enterprise AI agents, and his thesis that software costs are approaching zero while data becomes the only durable asset in the AI economy.

Cyber Warfare Reality vs Hollywood Fiction

"Once you start using missiles, most of these cyber elements are not really relevant because you would use cyber mostly to gather information and intelligence to prepare an attack" - Matt

A $20,000 drone attack on Amazon data centers proved more effective than multimillion-dollar cyber exploits, shutting down two AWS zones for 36 hours and affecting services from banks to Fortnite

Israel's cyber operations against Iran focused on psychological warfare through hijacked prayer apps and traffic light control for reconnaissance rather than destructive attacks

Iran's Underestimated Cyber Capabilities

"Many people from the military world, but also the intelligence community has been underestimating Iran capabilities exactly like people used to do with North Korea" - Matt

North Korea now has "some of the best hackers in the world" targeting financial institutions, demonstrating how underestimated nations can develop sophisticated capabilities

Iran's government shut down most internet access for users during the conflict, while disinformation campaigns proliferated across social media with AI-generated content

AI's Double-Edged Impact on Cybersecurity

AI tools like Claude are being used for vulnerability discovery in smart contracts and code assessment, making bug discovery "pretty good" according to recent Anthropic publications

"AI can also hallucinate, so even Darly or the CEO of Anthropic said it's definitely not in a state where it can be used for fully autonomous decisions" - Matt

Hackers are successfully jailbreaking AI systems to extract information, as demonstrated by the Mexican government system breach using Claude

Enterprise AI agents pose security risks when given "all permissions," creating Murphy's Law scenarios where "if something bad can happen because you give it access to it will happen"

The Coming SaaS Apocalypse and Data Economy

"The cost of building software is approaching zero" with tools like Claude Code, making it "hard to convince people that auditing software for security reasons is going to be more expensive than developing the actual software" - Matt

"Data is the only durable asset in the AI economy" as software becomes commoditized, leading to new business models around programmatic data access and API marketplaces

The Shopify CEO "can just rewrite an MRI software in one afternoon just to look at his back MRIs," illustrating the disruptive potential for existing SaaS businesses

Enterprise AI agents haven't emerged yet despite consumer adoption through tools like OpenAI's ChatGPT, with companies still "scared of being replaced for their jobs"

Future of Human-AI Code Interaction

"Humans are going to move towards creating markdown files as a programming language" with everything becoming "normal language" for machine interaction - Matt

Terminal and CLI interfaces are becoming the "natural interface for agents even for humans" as people move away from complex UIs toward direct command-based interaction

AI code quality improves with firm feedback: "if you say okay, like this piece of code is garbage, it's gonna understand better because you kind of give like a strong emotion" - Matt