How will we touch the robots?

New media and technology platforms usually start with content that imitates the format of their predecessors until a new native grammar emerges.

Early television borrowed heavily from live theater, with a single camera filming what were essentially plays or variety performances. As the medium evolved, television established its own grammar, transitioning from these static setups to multi-camera arrangements, close-ups, and other techniques now standard in TV production.

Likewise, the early days of the Web initially presented as a digital version of print media, with websites often resembling digital newspapers or magazines, differentiated solely by the hyperlink. But after some time, the interactive potential of the medium was realized and we entered the Web 2.0 era. You can draw a through line from early internet forums to social pioneers like Digg, Yelp and MySpace, to the modern behemoths we know today.

The rise of mobile technology followed a similar pattern. Early mobile apps frequently attempted to replicate desktop software experiences - the first iteration of the Facebook app was infamously a browser wrapper, much to Mark Zuckerberg’s chagrin .

As developers grasped the unique possibilities offered by mobile devices, truly mobile-native applications emerged. It took a little time, but eventually we got Uber, which took advantage of smartphones' GPS capabilities to create an on-demand marketplace for transportation, and Tinder, whose iconic swipe functionality was an innately touch-native mode of interaction.

This leads me to ask: what will AI-native user experiences look like? While the command line interface is a useful expedient, I think it’s unlikely that it’s the final form of how we will interact with models.

This is increasingly apparent as we enter the age of agents, which will take us beyond the prompt and response model familiar to any ChatGPT user.

For startups, novel user experiences are not just a nice to have - they have the potential to be a true differentiator. As I wrote last year , one of the paths towards establishing a new platform is a “different but better” mode of interaction that leans into the native advantages of a new technology. This is where 10x better products come from.

Superior user experiences alone will not create a moat. They can be copied by incumbents, and there is truth to the Andreessen adage that "the battle between every startup and incumbent comes down to whether the startup gets distribution before the incumbent gets innovation."

AI is non-deterministic, creative and unpredictable - my gut tells me that gifted designers will lean into these characteristics as they build the tools we will use to interact with it.

Game developers have been doing this for years. Long before we had “real” AI, we had games with artificially intelligent characters who could be interacted with in all kinds of strange, fun and intuitive ways.

Distinctive user experiences alone will not be enough for startups to ensure victory, but I think they can be a powerful tool for creating products that will delight users, accelerate adoption, supercharge organic distribution, reduce the cost of user acquisition and help startups build defensible brand and data moats.

If any of this rhymes with what you’re working on, we’d love to hear from you .