What If a Story Could Talk Back? Interactive Fiction Goes Mainstream

01 /The Concept

What if a story could talk back? And Then is exactly that: interactive story-games where users deliver their own lines – by voice or text – while an AI engine plays every other character in real time.

A few examples illustrate the range. One story begins with you sitting in a 24-hour diner when someone walks in carrying a bomb and threatens to detonate it. The police have been called, but they're still minutes away – and minutes may be all you have. Your only shot at preventing a catastrophe is to talk the bomber down yourself. So: start the conversation.

In another game, you have five minutes to convince someone to change their mind on a hot-button topic. Today's debate partner insists that Taylor Swift is wildly overrated. Yesterday's was a simulated AI lab researcher who believes AI is going to kill us all.

And Then was founded this year and just completed the a16z Speedrun accelerator. The startup hasn't settled on a monetization model yet – advertising, subscriptions, and pay-per-story are all on the table. It hasn't disclosed the exact investment from the accelerator, though Speedrun is known to deploy up to $1M, and that appears to be the figure here.

02 /Why It Matters

At the bottom of the And Then website there's an open call for creators to build their own story-games using the startup's existing – and forthcoming – tools. But why would they need outside creators at all? Can't the AI just generate the stories?

Surprisingly, this AI startup puts humans at the center. Every story and every character is conceived by a human creator; the AI's role is to "animate" those stories – turning them into live, responsive dialogues with users.

Because the AI is improvising from a human-crafted skeleton, every user gets a genuinely different version of the same story. It's the same dynamic as asking an AI chatbot the same question twice and getting different answers each time.

For And Then, that variability isn't a side effect – it's the core design feature, and the startup actively amplifies it. The same authored setting can generate an infinite supply of unique play-throughs, so even returning users encounter fresh plot developments and new arguments to deploy.

One of the stories was conceived and built on the And Then platform in four hours. The founder is convinced no AI could have come up with something equally gripping. In his view, AI can produce stories – but they tend to be formulaic. And despite rapid advances in AI capabilities, he believes humans will continue writing more compelling stories for a long time to come.

One could argue it might be "always" rather than "a long time." The bar separating a skilled human from AI will simply keep rising – and whatever AI can do will gradually stop being considered a mark of craft. Humans will have to keep stretching to stay above that bar. And this principle, of course, extends well beyond storytelling.

The same core insight – that only humans can invent extraordinary things, but AI can develop them just as brilliantly – drives a very different startup: Rocketable ([related review](/review/v-obshhem-sluchae-jeto-poka-fantastika-a-v-chastnom-vozmozhnost-na-milliard)), which graduated from Y Combinator this year and subsequently raised $6.5M.

Rocketable is building a "maximalist" software holding company – one person at the helm, an AI doing all the development work. The humans who conceive the products aren't even Rocketable employees. Instead, the startup acquires existing cloud services with at least two years of operating history, revenue between $500K and $2M, and a path to profitability – then hands them off to AI for further development.

03 /Opportunities

The broadest direction here is building startups around the "humans conceive, AI develops" model. The application spaces are vast.

In And Then, a human crafts the story skeleton; the AI turns it into a one-of-a-kind interactive dialogue experience, unique at every moment.

In Rocketable, a human builds a cloud service to the point of proven demand – and Rocketable acquires and scales it with AI.

On the Arcade platform ([related review](/review/hochesh-kupit-veshh-mechty)), a human designs jewelry or a custom rug; the AI translates that rough idea into a detailed production specification that artisans on the platform can execute. Arcade has raised $37M – $12M last fall and $25M in March of this year.

Madespace ([related review](/review/samoe-krutoe-sdelat-takoe-chemu-chatgpt-ne-konkurent)) launched an interior design platform in August: the human sketches the overall design concept, and the AI assembles a curated list of real furnishings that can be ordered directly through the platform.

So – in what domain would you build a platform where humans do the imagining and AI does the building and scaling?