Writings on AI
wm →
World Make is a newsletter about emerging technology, frontier AI, and the future of design. It is an experiment and a space to try to follow the rapid advances in the field and how to design for it. It features deep dives into product launches, best practices for design, UX/UI and evolutions in AI and agent interaction patterns. I try to find inspiration from lots of different types of project, and as the project evolves try not focus on the work of frontier companies.
At times it’s a bit technical, and focused around the evolution of agents, robotics and foundation world models - but I’ll always try and explain it in simple analogies to help grasp what it often complex. My main technological focus is designing for AI, robotics, agents and design. My longer term ambition with the project is to evolve the design patterns about Foundation World Models and Human-robot interaction. I find writing about these topics the best way to deeply engage with them and how to think about designing from them.I believe these are some of the biggest questions of our time - but let’s give it some personality and try to make it fun. As part of World Make is Weird History, where I’ll find unique design patterns or philosophy examples from everywhere to explore a facet of design and technology.
AI Agents
Agents are all the rage for 2025. Every frontier lab had agents on their roadmap for this year. Every week a new framework, mode of interaction and a new instruct model is released and evolves everything. But what are the best human-agent design practices? How are they being defined? How do we design for task and multi-agent systems?
Since 2021, I’ve been writing and thinking about AI agents and best practices for design for them.
Foundation World Models
World Models have always been an area of interest and research - models that can accurately model and represent the world around us through multi-modality pose new and unique opportunities to accelerate intelligence and build representational worlds for AI training.
Robotics
Human-Robot Interaction design is another area of passion and interest. Frameworks and interaction patterns for robotics interact with many different modes of AI - across World Models, Assistants, Agents, Tasks, Video, voice and motion. There's significant learning that can be leveraged from AI agent but the field is nascent and presents our most exciting challenges.
Design pATTERNS
Design Thinking →
Frame the Challenge and Problem
Before jumping into the design process, it's important to take a step back and really understand and frame the challenge. What problem are we really trying to solve here? Why? What's the value in focusing on this over other initiatives? This helps to align the team on a common goal.
Understand and Empathise with Users
Once we've understood the initiative to focus on the next stage is understanding users. Who are the users we're designing for? How can empathise and get a better understanding of what their current challenges are? What motivates them? Keeping the user at the centre makes sure your always designing with purpose.
Explore the As-Is and Current Scenario
Let's build and model what the current process and flow looks like. What does the current process look like? Why is it so challenging for users to navigate? Can we identify a variety of more granular challenges across the entire flow?
Reimagine, Sketch and Wireframe
Now we understand the challenge, process and users involved, let's start thinking about addressing these challenges and reimagining a new way to solve their problems. At this stage we start sketching, wireframing and re-invisioning what the process, app or flow could be.
Design and Prototype
After we have some initial mock-ups and lo-fi designs, we can start iterating in mid-fi and hi-fidelity prototypes. Vibe coding elements and parts of it, building prototypes in Figma and developing models and patterns that can be applied consistently throughout.
Reflect, Evaluate and Build
We've got a working front-end or hi-fidelity prototype - but how do we know how successful it's been in meeting our goal? User testing, researching and conducting interviews with users is the best way to understand the deeper nuances of what you've built, how you've built it and iterate on the future state and features. And repeat.
Follow Me On Substack
for weekly essays on AI and Design
