Introducing A New Creative Medium

What if you could start a conversation with your podcast? What if you could ask that hyper-specific question that keeps nagging you? What if the host answered, drawing on everything they know, including what never made the final cut? What if the episode kept growing after it was published — following your curiosity, until no two listeners heard the same show? What if, eventually, you stopped being a listener and became the third voice in the room, creating something completely new?

What if that were true of every form of media? Every book, every film, every song, every lecture, every game? Each one able to respond, adapt, change shape.

In such a world, the interplay among creator, work, and audience becomes part of the composition itself—an added dimension for the author to design.

We believe this new medium will be AI's defining creative possibility.

And none of this is science fiction. In fact, most of the necessary capabilities are already here.

The Necessary Pieces

First, today's large language models support long-context processing, multimodal reasoning, and low-latency inference, enabling increasingly interactive systems. Second, the infrastructure has matured in parallel: long-horizon memory, retrieval-augmented generation (RAG), and tool orchestration are now standard components of commercial platforms. Third—and most critically—user mental models are shifting. Daily interaction with conversational AI is normalizing participatory expectations, and users increasingly assume that digital content can be queried, redirected, and revisited.

The research community is converging on the same direction. Recent work in human–computer interaction treats the interface itself as adaptive material: documents that reorganize around a reader's goals, conversational systems that generate visual controls on demand, interfaces that restructure in real time as intent becomes clear.

Model capability, infrastructure, audience expectation, and research agenda are aligning on a single point. So the question is not whether this medium is possible—it's what it makes possible.

The Possible

Distribution has always imposed a condition on creative work: to reach an audience, expression must become a discrete, finished object. The resulting artifact may preserve language, images, sound, and narrative structure, but the creator's part in the communication effectively ends at publication. From that point, the work must speak for itself: whatever the creator would have clarified, expanded, or answered survives only as traces in the fixed form, and the labor of interpretation passes entirely to the audience. A published work receives no information from its encounter with the audience—confusion, objection, and curiosity—the artifact remains unchanged.

AI makes it possible to carry more of this communicative structure into the audience's experience. Creators can represent not only the finished expression, but also the knowledge, intent, perspective, and constraints that shaped it—and allow those elements to remain active during reception. When an audience member asks a question, expresses uncertainty, or introduces relevant context, that contribution can become an input to the experience itself. The work may clarify an idea, reorganize an explanation, surface supporting material, or return to an earlier point. Meaning is no longer confined to what was fixed at publication; it can continue to develop through interaction.

This does not transfer authorship from the creator to the model. The creator establishes the work's purpose, source material, perspective, expressive character, and conditions of participation. They do not script every possible response; instead, they author the system of intent, knowledge, and constraints from which responses emerge. The extent of audience participation is itself an authored property: some creators may permit guided exploration within carefully defined boundaries, while others may invite audiences to influence how an experience develops or contribute directly to it. Creators gain new dimensions of expression, and audiences gain the ability to follow their curiosity and engage more deeply with the creator's ideas.

The result is that the finished artifact is no longer the terminal boundary of expression. Publication becomes the beginning of an ongoing, authored exchange.

The Missing Creative Layer

But these outcomes won't emerge automatically from more capable models. Models provide technical capabilities, but capabilities alone don't constitute a mature creative medium.

Creators still lack established tools for representing communicative intent, organizing knowledge, defining what is fixed versus adaptive, setting behavioral boundaries, and determining the audience's degree of agency. They need ways to preview how a work might behave across different encounters and to evaluate whether its responses remain consistent with their intentions. Audiences need coherent ways to enter and leave interaction, preserve continuity, and understand which parts of an experience are authored, generated, or adapted.

A medium also requires conventions. Creators and audiences need shared expectations about authority, consent, attribution, memory, and control. The experiences themselves must feel native to the forms from which they emerge, rather than like generic interfaces attached after the fact.

That is where we come in.

We are building the creative layer through which intelligent works are authored and experienced. That means creating tools that let creators define a work's knowledge, intent, behavior, and boundaries. It also means building the destination where audiences encounter those works—and participate in them. The grammar of this medium doesn't yet exist, and we intend to develop it in collaboration with the people who will use it to create.

Our Starting Point: Podcasts

podchat is our first exploration of this thesis. Currently in beta, it already lets listeners move from a recorded podcast into a low-latency voice conversation about the episode. Grounded in the podcast's material, it can answer questions, clarify ideas, discuss arguments, and explore related subjects. With the host's permission, it can speak in their voice and conversational patterns.

The experience is explicitly presented as an AI extension of the host, not the host. Voice cloning requires the creator's consent. The purpose is not to convince listeners that they're having a private conversation with another person; it's to let a creator's work carry a bounded capacity for explanation and response.

Through using podchat, a podcast carries more than recorded audio into reception. It carries relevant context, an authored perspective, and a capacity for conversation. The listener can pursue an idea the moment curiosity arises while staying within the communicative environment established by the creator.

Podcasts are an ideal place to begin because they're already built around conversation. A host's voice, perspective, and explanatory style are central to the form, and listeners often develop lasting relationships with the people and ideas behind their favorite shows. Audio can also run at low enough latency for interaction to feel natural.

Different podcasts will require different knowledge, behaviors, boundaries, and degrees of participation. Producers therefore need to be co-developers of the form. By working with them, we hope to learn which kinds of interaction deepen a podcast, which compromise its intent, and what creators need in order to exercise meaningful control.

Where We're Heading

By building alongside podcast producers, we can begin developing the authoring tools, interaction patterns, and creative conventions required for adaptive, creator-directed experiences. We'll then extend those capabilities to other media, each of which will discover its own expressive uses for them. We're extremely excited about this—and the only thing we can reliably anticipate is that creators will use memory, adaptation, simulation, and co-creation to invent forms that are un-anticipate-able!

We are technical storytellers working at the boundary between creative practice and computational systems. We believe this medium will be developed by people who understand both the mechanics of these systems and the human purposes they should serve.

AI will not only change how existing media are produced. It will expand the forms through which people create, participate, and communicate.

If you believe what we believe, we would love to hear from you: support@podchat.dev