Research
Memory & Agency: What Should an Agent Remember?
A useful agent memory keeps the evidence, context, and limits behind a lesson so it can be tested before being used again.
The Geometry of Intention: What Behavior Can Tell Us About Human Goals
Clicks and searches reveal what people do. Context can help us infer why, but a useful system must remain open to being wrong.
Your Marketing Agent Needs a World, Not Another Prompt: Building Marketing Agentic Systems the Palantir Way
A marketing agent needs a reliable way to connect a request to real campaigns, approved requirements, evidence, permissions, and outcomes. Palantir's Ontology offers one useful model for that operational layer.
Building to Learn: What Three Weeks of Prototyping Taught Us About Agentic Marketing Optimization
A three-week prototype explored whether simulated product-intent scores could help improve AI shopping visibility. The work exposed a useful boundary: a lab can assess semantic fit, but only external evidence can test discovery or business impact.
Who Owns the Pixel? ChatGPT and the Repricing of Advertising
ChatGPT ads put conversational context and advertiser outcome data on different sides of a measurement boundary. That changes what brands and agencies can know, test, and credibly claim.
The Empowerment Imperative: What Should Agentic Marketing Optimize?
Marketing systems can optimize measurable actions. This essay asks what else they should protect, how to turn human agency into a design question, and what evidence would test the proposal.