Journal: May 2025

Journal for May 2025

Monthly retrospective

Created a project template based on Cloudflare Workers + Cloudflare D1 + Durable objects + SvelteKit, and continued working on two LLM projects in May, following on from April.

Articles written in May:

• Premature death of UI design
• Agentic coding experiment 2

Journal

2025-05-25

There were all sorts of announcements: Google I/O, Google Cloud Next, Microsoft Build, Figma Config, Code with Claude, and a surprise announcement from OpenAI (since OpenAI DevDay has traditionally been in the fall, OpenAI sometimes makes surprise announcements in May to take the wind out of other companies’ sails).

It’s overwhelming.

2025-05-24

Side project idea: There are so many companies to boycott that I keep forgetting which ones to boycott and why, can’t manage my shopping, and our efforts get scattered. Solution: set topics I care about, then see, by sector, the top-ranked company (focus the boycott) and bottom-ranked company (focus purchases) for causing recent, frequent, and major problems.

2025-05-15

Had a meeting, so visited a café called “Oriu” near Naebang Station. Pretty and quiet.

2025-05-14

Yesterday Google released Material 3 Expressive, a revision of its design system Material Design 3.

• Greater emphasis on emotion
• Improved usability across all ages (especially older people)
• Improvements based on scientific research, but without the “engineering-centered” approach of the past

The third point stands out, because Google’s engineering-centered design culture caused a major controversy in 2009. When Google’s design team lead left for Twitter, he pointed out problems along these lines: “I’m the design team lead, but I can’t decide for myself which blue border to use or how many pixels the corner radius should be; I have to adopt the results of A/B tests run by engineers.” Engineering-oriented optimization was excessively constraining designers’ creativity.

It was striking that this time Google explicitly mentioned the 2009 incident, saying it had avoided repeating that mistake.

“

Material 3 Expressive was born out of research. But rather than the old ‘41 shades of blue’ approach that delegated design decisions to data, it emerged through collaborative inquiry spanning research, design, and engineering. (Material 3 Expressive was born out of research — not in the 41 shades of blue kind of way, which delegated design decisions to data, but in a collaborative inquiry spanning research, design, and engineering.)

2025-05-10

Studied Cloudflare Workers a little. I now know how to connect SvelteKit and Durable objects, which lets me use the Agent API.

2025-05-07

Built a Generative Wikipedia prototype.

2025-05-05

Trying out Agentic coding experiment 2.

2025-05-03

I’ve occasionally heard claims for about fifteen years that there is nothing new left in HCI or UI, and lately I hear them more often. There is also the claim that so-called “thin AI wrapper” products, which add a little UI to an LLM API, are not competitive. VCs have been going on about this for the last few years. Another claim is that the AI era means we no longer need UI designers or frontend developers. I disagree with all three interconnected claims. Wrote up these thoughts: Premature death of UI design.