
Sign up to save your podcasts
Or


AI is making it easier to understand, recreate, and investigate software. But who decides when those capabilities are off limits?
In this episode of Gradient Descent, Chris Riordan, Brett Lucas, and Brett “OB” Haupt explore the tension between protecting intellectual property and enabling independent security research. When the same tools can help replicate a commercial product or uncover a hidden vulnerability, where should AI providers draw the line?
The conversation moves from software’s potential “Napster moment” to the value of proprietary data, the role of open-weight models, and the risks of selective enforcement. What happens to innovation, competition, and cybersecurity when access to powerful tools depends on who gives you permission?
What happens when an AI agent has to make a decision that its instructions don't account for? Should it follow orders, refuse, or ask for help?
In this episode of Gradient Descent, Chris Riordan, Brett Lucas, and Brett Haupt (OB) welcome Brock Christoval, CEO and founder of Progredi Systems, to explore constitutional AI and the challenge of governing autonomous systems in high-stakes environments.
From military rules of engagement to AI agents that must operate without human oversight, the conversation examines how models interpret authority, navigate conflicting instructions, and determine when to stop and ask.
Brock also shares how Progredi Systems is approaching mission-oriented AI governance, including the evaluation of agent behavior, decision-making, and accountability.
Can we build AI systems that do more than follow instructions? And who gets to decide the principles they follow?
Join the conversation as the team explores the space between AI capability, autonomy, and responsibility.
AI can write code faster than ever. But what happens when an entire development team starts working at that speed?
In this episode of Gradient Descent, the RavenTek team gets practical about AI-assisted software development, sharing lessons from a recent project that generated 933 commits, 37 development collisions, and plenty of unexpected challenges.
From conflicting code changes and runaway documentation to the future of Agile development, the conversation explores what changes when AI dramatically accelerates the pace of building software.
Can traditional development practices keep up? How much oversight do AI coding agents really need? And does faster development actually mean better software?
Join the team for a candid conversation about what worked, what didn't, and why human judgment matters more than ever.
What happens when the technology transforming how we search, write, code, and make decisions depends on enormous amounts of data, compute, and capital?
In this episode of Gradient Descent, Chris Riordan, Brett Lucas, and Brett Haupt introduce a new ongoing conversation about “authoritarian AI,” the idea that the economics and architecture of modern AI may naturally concentrate power in the hands of organizations with the most data and resources.
The team breaks down how large language models are built, why open models matter, where bias and guardrails enter the equation, and the tension between AI safety, regulation, competition, and openness.
This is only the beginning. As AI becomes more embedded in everyday life, who controls the models, the data, and the rules governing them will become increasingly important.
AI adoption is accelerating, but are organizations securing it as quickly as they are deploying it?
Rob Chee, Federal Solutions Architect at Zscaler, joins the Gradient Descent crew to explore how AI red teaming, guardrails and zero trust can help organizations identify vulnerabilities before attackers do. The conversation covers prompt injection, jailbreaking, model testing, agentic security and the challenge of protecting AI applications without slowing innovation.
The group also examines the risks of AI-generated software, the changing role of human judgment and why no single security layer is enough.
AI feels cheap. A monthly subscription, a few prompts, maybe a coding assistant. But that price doesn't necessarily reflect what AI actually costs to run.
In this episode of Gradient Descent, Chris Riordan, Brett Lucas, and Brett Haupt unpack the economics behind AI, from tokens and reasoning models to local LLMs, data centers, energy demand, and the growing cost of agentic workflows.
They explore why today's AI pricing may not be sustainable, whether smaller and specialized models offer a better path forward, and what happens when organizations have to start accounting for the true cost of the AI they've built into their workflows.
Because the question isn't just what AI can do. It's what happens when the bill comes due.
Fresh from Black Hat USA 2026 in Las Vegas, Brett Lucas and Brett Haupt join Chris Riordan to unpack what they saw across the cybersecurity industry.
AI was everywhere, but that was part of the problem. When every product claims to use AI, the more interesting questions are what is happening under the hood, what decisions are being handed to LLMs, and whether those systems are actually explainable, auditable and worth the cost.
The team digs into AI-powered security products, the difference between using LLMs as communicators and decision-makers, token-based pricing, vibe-coded interfaces, Black Hat paranoia and the technologies that actually caught their attention.
The biggest takeaway? The industry may finally be moving past “just use more AI” and back toward thoughtful architecture, guardrails and fundamentals.
Everyone talks about AI. Almost nobody talks about how they actually use it.
In this episode of Gradient Descent, Chris Riordan, Brett Lucas, and guest Guy Waybright compare their personal AI workflows, from governance files and context management to coding agents, testing, and multi-model orchestration.
The surprising takeaway is that there isn't one right way to work with AI. The best workflow is the one that matches how you think.
Whether you're building software, managing projects, or just trying to get more out of ChatGPT, this conversation offers practical ideas you can apply today.
What happens when artificial intelligence becomes both the attacker and the defender?
As AI accelerates cyber threats, organizations face a difficult question: should they fight AI with even more AI, or double down on proven security principles like Zero Trust?
In this episode of Gradient Descent, Chris Riordon, Brett Lucas, Brett Haupt, and Ryan Sheehan debate where AI belongs in modern cybersecurity. They explore agentic AI, prompt injection, AI-powered vulnerability discovery, offensive automation, and why human oversight may become even more important as machines get faster.
Rather than predicting doom, the conversation focuses on how organizations can adopt AI responsibly while protecting the systems that matter most.
Topics include:
Everyone is celebrating how AI is accelerating software development. Far fewer people are asking what happens when security can't keep up.
In this episode of Gradient Descent, the team introduces the idea of Apocaloptimism. It is the belief that AI has incredible potential while recognizing that its rapid adoption may create serious challenges before things stabilize.
The conversation explores two trends moving toward each other. AI is making it faster than ever to build software, while AI-powered offensive capabilities are lowering the barrier to sophisticated cyberattacks. What happens when those two curves finally intersect?
Join the discussion as the team explores:
AI-assisted software development and "vibe coding"
Why software quality and security may be falling behind AI adoption
How offensive AI is changing the cybersecurity landscape
Zero Trust, governance, and responsible AI implementation
Whether we've already reached the tipping point, or if the biggest disruption is still ahead
Not everyone agrees. That's what makes the conversation worth having.
From the publisher's feed
Gradient Descent is a podcast for people asking better questions about emerging technology.
Hosted by RavenTek's Chris Riordan, Brett Lucas, and Brett Haupt, the show explores artificial intelligence, cybersecurity, leadership, and software development through thoughtful conversations that challenge assumptions rather than reinforce them.
We aren't interested in repeating the latest buzzwords or predicting that every new technology will save the world or destroy it. Instead, we explore the space between the hype and the fear, sharing practical perspectives for leaders navigating an increasingly complex world.