Last night, after leaving the hackathon, I started to feel as though time was moving more slowly than usual, and my brain felt strangely agitated. The sensation intensified toward bedtime, and I ended up unable to fall asleep at all. After investigating, I realized that several cans of “sparkling water” I had consumed during the event actually contained a large amount of vitamins and stimulants, including caffeine and green tea extract. In effect, they were closer to Red Bull or Monster energy drinks than to ordinary sparkling water. I should have paid more attention to the nutrition labels, but I was too preoccupied with the hackathon to do so.
2026-10-12
2026-10-11
MongoDB hackathon
This hackathon drew a much larger crowd than the previous one I attended, with hundreds of participants and 93 projects submitted in total. Unlike the previous event, the judging emphasized topic impact and live presentation more than technical execution. Since the theme focused on harnesses and databases rather than motion and graphics and I also ran into hardware issues during my presentation (screen sharing and a broken mouse), I did not perform as well as last time. Even so, the project ranked 10th out of 93, was completed well ahead of schedule (with most of the time spent on demo production rather than on the actual development), and I had a great time on Sunday in a wonderful venue with nice food and drinks.
See this repository for more information about the project, which implements a harness and database backend for my earlier DrawTalk frontend.
https://www.instagram.com/kub1ai/p/DeZxSg2FoVv/
2026-10-08
Temporary Body by Jerron Herman
In this participatory art project, the artist first invited the audience to create decorative adornments before a solo dance performance among the audience. I enjoyed making a few quick headdresses and immersing myself in the performance. My interpretation is that these adornments and dances act as positive beautifications for the artist, who appears to be gay and struggles to move the left side of his body.
https://www.instagram.com/kub1ai/p/DeUkRkommBG/
https://www.behance.net/gallery/256925537/Temporary-Body-by-Jerron-Herman
2026-09-30
Runway hackathon
For the first time in my life, I participated in a hackathon, hosted by Runway, whose products I have long admired. I especially like their motion brush feature, though it seems to have disappeared after Gen-2.
For this hackathon, I wanted to revive the motion brush as a direct manipulation tool that could complement other input modalities such as text prompts, keyframe images, and guide videos. See this repository for more information, including code, process, and demo videos.
Overall, I found this to be a very satisfying experience: an intense 6-hour event in which I could fully focus on ideation, design, implementation, experimentation, and demo production. Initially, I thought I would be the only participant over the age of 20, but the demographic turned out to be more diverse than I expected. I also got to see some very interesting demos from other participants.
https://www.behance.net/gallery/256531401/RunwayHackathon
2026-09-20
Hand-building
Hand-building seems more forgiving than throwing, as we can add and remove pieces of clay to the model gradually.
https://www.instagram.com/p/Ddh_Jg8G0Zu/
https://www.behance.net/gallery/256288399/Handbuilding
2026-09-13
Kusama: Infinity
I watched this documentary on hoopla about Yayoi Kusama’s life and art. It chronicles her difficult childhood, struggles with mental illness, and her resilient journey from decades of isolation and obscurity to eventual acclaim and influence on the art world.
The documentary is visually compelling and deeply inspiring. I highly recommend it to anyone interested in art, psychology, and the power of perseverance.
2026-09-02
Interface world model
From TLDR newsletter about Runway’s Solaris:
Solaris is an Interface World Model. It can generate real-time interactive interfaces frame by frame as users interact with it. Solaris handles rendering and interactions jointly, generating every frame and every response to user input, eliminating the need for an intermediate representation. The model opens up new ways of building websites, apps, and other online interfaces. It can also be used to train agents in much more dynamic environments.
Very cool demos!
The basic idea is to use a MLLM to generate prompts from user inputs across multiple modalities (including texts, direct manipulation, etc.) and the current image frame content, and use the prompts to generate the next frame via a fast video diffusion model.
The article heavily emphasizes the contrast with approaches that generate an intermediate representation (like code) which can be too lossy for more complex images (like natural photographs rather than simple layouts).
Because the “interface” can be generated entirely on the fly at the OS level, it can replace the traditional app-based approach and allow for more dynamic and personalized interfaces that orchestrate underlying modules/tools to meet the user’s needs and preferences in real-time.
How to connect this interface world model with real computation substrate can be an interesting next step (e.g., via an agentic orchestration layer that can translate the interface world model into actual computation and data retrieval).
2026-08-31
Great collaboratorg
While writing my research statement recently, I realized how much I owe to Rubaiat Habib Hazi for my work over the past decade. He has been a great collaborator and my number one co-author on my Google Scholar profile.
When I expanded from graphics to HCI, thanks to my first official PhD student, Jun Xing, who interned with Rubaiat at Autodesk, Rubaiat provided timely guidance on HCI and content creation throughout our collaborations. When I joined Adobe, I worried that I could no longer work with him, but fortunately, he joined me just a few months later. We continued our collaborations over the years, resulting in 25 publications (including 5 best paper or honorable mention awards) and product features in Fresco Motion and generative motion graphics.
I am grateful for his support and friendship, and feel lucky to still be able to work with him after all these years.
2026-08-28
Artisans and startups
Before the industrial revolution, artisans were the main providers of goods and services. They often had to be skilled in multiple areas, such as design, production, and sales, to run their own small businesses.
The industrial revolution led to the rise of large corporations and the division of labor, where different roles were specialized and separated. Artisans were replaced by factory and office workers, whose skills were often limited to specific tasks within a larger production process.
The advance of AI and automation has the potential to bring back some of the artisan spirit, as individuals and small teams can leverage technology to create and distribute products and services without the need for large teams. This could lead to a resurgence of small businesses and startups, where individuals can take on multiple roles and have more control over their work.