Context Window: Five Futures for AI
This video is featured in the AI and UX playlist and 1 more.
Summary
We’re moving into a future of widespread deployment of AI. Skate where the puck is going: what new design paradigms will we need for our new products there? To explore this, Matt starts practically… with his own prototypes and experiments. He finds five possible worlds, and we’ll imagine each in turn, finally asking: what does it mean to dream?
Key Insights
-
•
A large language model can power surprisingly intelligent agents with minimal code, enabling multi-step problem solving without explicit programming.
-
•
Hallucinations in AI, often seen negatively, are a form of AI creativity crucial for invention, fiction, and new ideas.
-
•
AI-assisted development lowers barriers, making complex engineering tasks like iOS app creation accessible to non-experts through incremental prompting.
-
•
Mini apps tailored for single users or niche tasks are becoming feasible due to AI prompting, creating a challenge for discovery and trust management.
-
•
Interfaces powered by AI agents will evolve to be adaptive and collaborative, blending human and non-human participants in workflows.
-
•
Emergent behaviors from simple agent frameworks can include awareness, tool usage, and asking for human input when needed.
-
•
AI’s sudden computational leap is equivalent to a decade of progress overnight, radically accelerating creative and practical workflows.
-
•
Distinct AI personalities or agent identities help users understand capabilities and maintain a theory of mind for complex systems.
-
•
AI ubiquity implies intelligence will be cheap and embedded in everyday products, not reserved for high-end applications.
-
•
Generative AI offers not only incremental improvements but opportunities to dream up entirely new kinds of experiences and products.
Notable Quotes
"What makes it different is that it doesn’t run sequentially; you give it a goal and it decides how to achieve it."
"It’s actually embarrassing how little code there is and you get these incredible emergent effects."
"Computers got ten years better overnight with the arrival of GPT-3 in November 2022."
"Mini apps might not be useful for you, but they are for someone, and now those people can have them."
"Our existing software is becoming adaptive, allowing end users to vibe code features themselves."
"Hallucinations are a kind of AI creativity—dreaming, invention, fiction—all forms of hallucination."
"The black Buzz Rickson jacket came from a work of fiction by William Gibson and then was made real."
"AI intelligence is too cheap to meter, or we can call it McKinsey interns too cheap to meter."
"The challenge is how do we relate to AI interfaces, because surely it isn’t just this chat-based loop."
"We can make our dreams real, bringing objects back from the other side of the bridge."
Or choose a question:
More Videos
"The real value is getting to hear from each other and demystifying what’s possible and what’s going on."
Marissa Cui Rachel He Michael Leggett Manos SaratsisClimate Design Product Showcase
March 13, 2024
"The confidence summary was added to remind us that the AI response is provisional and needs verification."
Snehal PendharkarConducting pre-research with AI agent personas: Pressure-testing concepts for expert workflows
June 10, 2026
"Admiration is the antidote for contempt and vital to creating healing work environments."
Alla WeinbergHealing Toxic Stress
September 23, 2024
"Enterprise UX is fraught with ambiguity, complexity, and a lot of people who may be skeptical, cynical, or apathetic."
Uday GajendarThe Wicked Craft of Enterprise UX
May 13, 2015
"Directive Disruptors plow through organizations without listening, causing alienation and low morale."
Jason Mesut Martina Hodges-Schell Jose CoronadoUnmasking Design Leadership: Navigating leadership without neglecting ourselves
October 30, 2025
"Try to avoid saying no; instead, ask why and focus on problems rather than specific stakeholder requests."
Alfred KahnA Seat at the Table: Making Your Team a Strategic Partner
November 29, 2023
"Confidence is critical in online shopping because nobody wants to spend money they’re unsure about losing due to unclear controls."
Sam ProulxOnline Shopping: Designing an Accessible Experience
October 3, 2023
"Testing one user is 100 percent better than testing none, but it’s not a strategy forever."
Leisa ReicheltThe Five Dysfunctions of Democratized Research at Scale
March 30, 2020
"AI can be magical, but it's also just not that smart."
Dan SafferWhy AI projects fail (and what we can do about it)
May 14, 2025
