An LLM, by itself, cannot be truly agentic. Same for swarms, teams, workflows, or “multi-agent” systems. If the LLM drives everything, it’s not agentic. LLMs can be useful appendages, but not a sound foundation.

arxiv.org/abs/2505.03434

Technology is enabling, but also constraining. Choosing a tool means trading flexibility in one dimension for scale in another. Customer engagement is no different. Campaigns, in particular, aren’t neutral abstractions—they’re design choices with real consequences.

We throw around terms like “agency” in AI, but the word itself lacks definition. That problem predates AI. Philosophers have debated it for centuries without consensus: Do instincts count? Are reasons different from causes? Is planning the same as action? Does coercion erase it?

*How* a system navigates the explore/exploit tradeoff is more important than the claim that it does so. If you’re trying to build or evaluate an agentic system, here are the questions I think you need to be able to answer.

A/B tests, predictive models, bandits, and agentic learners look similar but behave differently. Think of them as investing strategies: each makes choices under uncertainty, but with different assumptions about markets, change, and context.

We just raised an $18M Series A, led by Theory Ventures. I’m proud of what we’ve built.

www.forbes.com/sites/davidp...

This. Generative models are procedural memory machines. Procedural memory is a really important part of human cognition, but it's not the only part - not by a long shot. There are so many other insertion points if we want to try to get a computer to emulate cognition.