AI-Curious with Jeff Wilser

What happens when your next customer is not a person, but an AI agent shopping on their behalf? The way we buy things online may be about to change a lot faster than most people realize.
In this episode of AI-Curious, we talk with Chi Zhang, cofounder and CEO of Kite, about agentic commerce and the infrastructure that could make AI agents true economic actors instead of just helpful assistants. We explore what it means to let an agent not only research flights, groceries, APIs, or consumer goods, but actually complete the transaction safely. Along the way, we unpack why Chi describes Kite as the “Stripe for agents,” and why this shift could force businesses to rethink who they are really selling to.
We also dig into the hard part: trust. If an agent is going to spend money on your behalf, how does a merchant know that your agent is legitimate, authorized, and not a scam wearing your face? We get into the identity, verification, authorization, privacy, and infrastructure layers that make agentic payments possible, and why those pieces matter just as much as the agents themselves.
This conversation also looks at why stablecoins and programmable money may be especially well suited to this future, particularly for micropayments, API access, and machine-to-machine commerce where traditional card rails are too expensive or clunky. More broadly, we talk about what happens when AI agents start doing more of the comparison shopping, checkout, and transaction work that humans used to do themselves.
Guest
Chi Zhang — Cofounder and CEO, Kite
Check out Kite AI
Follow AI-Curious on your favorite podcast platform:
Apple Podcasts
Spotify
YouTube
All Other Platforms
For anyone interested in Jeff’s AI Workshops for their company:
Reach out directly at [email protected]

In a year when the frontier AI labs are spending tens of billions on training runs and scaling their teams into the thousands, one of 2026's fastest-growing agentic AI companies has exactly one employee.

His name is Arthur Vandelay. His company is Mojo Agentic AI. And his revenue, with zero outside funding and a Fortune 100 client roster, is on track to clear $30 million this year.

In this episode of AI-Curious, Arthur walks us through how he did it — a master class in vibe coding, the agentic harness architecture he built from scratch, and the human-in-the-loop and governance disciplines that let a one-person shop sell credibly into the Fortune 100. We get into where the agentic AI hype is real and where it isn't, why the model is no longer the bottleneck, and what most enterprises are still getting wrong about putting autonomous agents into production.

If you're a builder, an operator, or an executive trying to figure out where agents actually fit in your business, this is one of the most concrete glimpses we've gotten this year into what a working agentic AI company looks like at scale.

Guest
Arthur Vandelay — Founder and CEO, Mojo Agentic AI.

What if the future of AI is not just better text, better image, and better video models stitched together, but something closer to a unified mind?
In this episode of AI-Curious, we talk with Caroline Ingeborn, COO of Luma AI, about the company’s bet on “unified intelligence” and why that may be a fundamentally different path toward AGI. We explore why Luma believes training across modalities together, instead of building separate models and bolting them together later, could unlock more natural reasoning and much more powerful creative tools. We also get into Luma’s latest release, Uni 1.1, a thinking image model trained on both image and text, and what that means for editing, image composition, and creative control.
We also look at how this is already changing real creative work. From agencies showing up to pitch meetings with finished videos already made, to Japanese animation studios using AI to move faster without sacrificing quality, we discuss what happens when creative teams can build worlds instead of generating image by image. Along the way, we talk about Luma’s creative agents, how they help turn scripts and briefs into characters, storyboards, and scenes, and why the goal is not to replace human taste, but to multiply it.
This conversation also goes deeper than tools. We talk about AI slop, human performance, visual communication, the future of agencies, and why the best creators may be the ones who learn to work with these systems earliest and best. If multimodal intelligence is real, what does it mean to build machines that think more like we do, and what does that change for storytelling, creativity, and work itself?
Guest:
Caroline Ingeborn — COO, Luma AI
Follow AI-Curious on your favorite podcast platform:
Apple Podcasts
Spotify
YouTube
All Other Platforms
For anyone interested in Jeff’s AI Workshops for their company:
Reach out directly at [email protected]

What if the future of AI is not bigger models in bigger data centers, but smaller ones running quietly on the devices you already use every day?
In this episode of AI-Curious, we talk with Jeffrey Li, COO of Liquid AI, about why the next phase of AI may depend less on giant cloud models and more on small, specialized models that run directly on phones, laptops, cars, and other edge devices. We explore the case for on-device AI, why large models are only part of the story, and how companies should think about speed, privacy, cost, and real-world deployment as AI moves from experimentation to everyday products.
We also dig into the economics behind this shift. Along the way, we discuss why cloud-based AI can break down when every query has to travel to a data center, why enterprise ROI gets harder as AI subsidies fade, and why many real-world use cases do not need a giant model capable of doing everything. Instead, they may need a smaller, more tailored system built for a specific task, domain, or device.
We also get into Liquid AI’s research roots at MIT, the origins of liquid neural networks, and what it looks like to bring production-quality AI into places like Mercedes vehicles and e-commerce systems. This is a practical conversation about the future of edge AI, specialized models, privacy-preserving AI, and what happens when intelligence moves closer to the user.
Guest
Jeffrey Li — COO, Liquid AI
Follow AI-Curious on your favorite podcast platform:
Apple Podcasts
Spotify
YouTube
All Other Platforms
For anyone interested in Jeff’s AI Workshops for their company:
Reach out directly at [email protected]

It’s happening everywhere. And no one’s really talking about it. What happens when your employees are already using dozens of AI tools your company never approved?
In this episode of AI-Curious, we talk with Rick Caccia, co-founder and CEO of Witness AI, about the rise of “shadow AI” inside enterprises and why it has become one of the biggest practical challenges in AI adoption. We explore how employees, often with good intentions, are quietly using ChatGPT, Copilot, and thousands of other AI apps to do their jobs faster, sometimes with sensitive data that should never leave the company.
We also dig into what happens when that behavior scales. From customer support teams pasting financial information into AI tools, to marketers uploading customer lists, to developers sharing source code with external models, we look at the real security, compliance, privacy, and cost risks companies are now facing. We also discuss why this problem gets even harder with AI agents, which can take actions, access systems, and create new forms of risk far beyond a simple chatbot prompt.
Along the way, we talk about prompt injection, jailbreaks, token costs, insider risk, enterprise governance, and how leaders can build an AI strategy that enables productivity without creating chaos. This is a practical conversation for anyone trying to understand how AI is actually being used inside organizations right now, and what it takes to manage that responsibly.
Guest
Rick Caccia — Co-founder and CEO, Witness AI
Follow AI-Curious on your favorite podcast platform:
Apple Podcasts
Spotify
YouTube
All Other Platforms
For anyone interested in Jeff’s AI Workshops for their company:
Reach out directly at [email protected]