Ep 18. The Ad Tech Plumbing That Makes Agentic Advertising Actually Work, with Brian Tomasette

Mark Zuckerberg says just tell AI what you want to sell and it handles the rest. Brian Tomasette — 25-year ad tech veteran, former Amazon DSP product leader, and current PhD researcher — says that's high school physics without friction. The business rules, privacy constraints, and data ownership that make advertising actually work don't disappear because the model is powerful enough. They just need better tools to operate them.

This week, Miles sits down with Brian to get into the real plumbing behind agentic advertising — what's actually being built today, how far off the utopian vision really is, and why the tools aren't going away anytime soon. Brian is currently researching neuro-symbolic AI at Drexel University and contributing to the IAB Tech Lab's AAMP framework, which is building the buyer and seller agent infrastructure that could eventually reshape how media gets bought and sold.

They break down why AI agents will use DSPs more optimally rather than replace them, what the IAB Tech Lab's buyer and seller agent SDKs actually do, and why Andrej Karpathy is right that it's not the year of agents — it's the decade. Brian also introduces the "AI vampire" phenomenon — the way vibe coding becomes as addictive as TikTok and drains you faster than actual genius work ever could — and why the right move is sometimes just closing the laptop and going for a hike.

Links from this episode:

AAMP — Agentic Advertising Management Protocol https://iabtechlab.com/standards/aamp-agentic-advertising-management-protocols/

IAB Tech Lab Buyer Agent Documentation https://iabtechlab.github.io/buyer-agent/

IAB Tech Lab Seller Agent Documentation https://iabtechlab.github.io/seller-agent/

Andrej Karpathy: "It's Not the Year of Agents, It's the Decade of Agents" https://www.youtube.com/watch?v=BlVnGXEzFow

The AI Vampire — Steve Yegge https://steve-yegge.medium.com/the-ai-vampire-eda6e4f07163

Beads — Steve Yegge's Claude Code library https://github.com/steveyegge/beads

Neuro-Symbolic AI Research Paper https://arxiv.org/abs/2510.05486

HAZ Lab at Drexel — Brian's Research Lab https://zharry29.github.io/hazlab/

Ready to level up your AI skills? Join the U of Digital AI Accelerator. Now enrolling for May | https://uof.digital/ai/accelerator/

New episodes every Monday. Get your AI Edge here.