Some quick notes after spending a year on applications of AI and thoughts on what's next ✌️ (originally posted on X, edited for clarity)
Agents: We're in the GPT3 era of agents. Most companies optimize for shiny demos, but few can handle hard task completion over multiple steps with enterprise-grade accuracy. That changes in 2025.
Enterprise Adoption: Founders underestimate how every enterprise is finally looking to buy AI applications. This buying opportunity happens once in a generation. Usually you're pulling teeth to get adoption. If you're not building AI-native applications, you need 10x conviction on why not.
Experiences: There's a shocking lack of experimentation in customer experience. Most build chatbots on basic RAG or stitch together no-code workflows. Massive opportunity for designers to leverage voice, image, and text to build entirely new experiences.
Boring Verticals: Greenfield industries remain untapped for even basic products. While everyone builds for engineers, lawyers, doctors, and support teams, entire industries wait for innovation.
Robotics: Robotics is finally having its moment. Perfect timing: labor shortages, efficiency demands, vision models, Starlink connectivity. If I were starting fresh, this is where I'd focus.
Context: Personalization remains underutilized. Every pixel will be generated on-the-fly soon—not just images and games, but every interface. Why constrain UIs to bounded boxes when you can prompt what you want?
Incumbents: Building application companies feels risky in the shadow of large model providers. But with models stabilizing and test-time computation emerging, the next 12 months offer massive opportunities. Vertical integration will be key for customer lock-in.
Consumer: The future isn't billion-user monolithic apps. It's constellations of focused applications with shared identity layers. True for both consumer and enterprise.
Agent Infrastructure: We need infrastructure for agents. As non-human identities (NHIs) become primary internet users, companies need frameworks to handle them.
Thought → Application: Products that enable creativity and bridge thought to pixels will flourish. In an era of abundance, meaning becomes scarce. Enable meaning-making, and you win.
Finally, thank you all for reading and sharing your thoughts this year. Writing has connected me with incredible people I had no business meeting. Try it yourself—you might be surprised. See you in 2025 👋
Thanks for this concise summary. As a late (or still early?) entrant in this space, I am looking forward to learn more about AI through the rest of the year and beyond!