Back to first principles
Weeknotes from the 11 April 2025
Things I did #
At work #
The last two weeks have been really productive. The team I'm working with is changing shape, with temporary gaps needing to be filled. This opportunity means I get to pick up some interaction design work, which is great.
Prototype migration #
One task I've been looking at is moving Figma designs into the GOV.UK prototype kit. The main reason for this is that there's more scope for iteration on a specific product, and as we move into user research getting granular with interaction will be key.
Moving designs from one platform to another is quite laborious and could easily sink a few days, perhaps close to a week. You could argue it's "busywork," and I could list a few other activities that might be better value for the UK taxpayer. However it was needed to be done, but I looked into ways to automate this task.
I decided to use Microsoft Copilot to generate the HTML from the Figma designs, as the department is using it on trial. I made sure the images didn't have anything sensitive and the designs were generic.
I followed these simple steps:
- Export screenshots
- Create empty screens in the prototype kit, with the head and footer
- Add the images to Copilot, with a prompt: Please convert this into GOV.UK prototype kit syntax using the table component, ensuring it matches the design standards for a GOV.UK screenshot.
- Copy outputted HTML from Copilot into the blank template files in VS code
- Tested the output locally and made any adjustments
The good
This worked really well, and I didn't need to specify the component, and in many instances, it worked well without it. Copilot also did a good job providing hidden text for screen readers.
This saved hours of time and does just what I want from AI, automating mundane tasks. Although I could use Copilot to create designs for user pain points, I really don't want AI to do that. It's by far one of the most enjoyable aspects of my work.
The bad
When giving Copilot full screenshots, the results sometimes are a little less consistent, occasionally making designs full width and forms and input fields very wide.
For better results, breaking apart more complex designs into components works well.
The "Check your answers" screen also had mixed results so splitting out the sections of the page worked much better.
Prison process mapping #
I also enjoyed some mapping of processes and operating models in a specific prison that is trialing one of our products. It's working really well in that prison, so we are trying to visualize what good looks like and see how scalable the processes are.
In-person planning #
Had a great day in Sheffield planning for the next year of work, taking pain points, rephrasing them as opportunities, and mapping them against impact and confidence.
We ended up with a loose roadmap and a growing backlog of activity.
At play #
Bikablo #
Outside of work, I took an intro course to Bikablo. Really enjoyed it and hopefully it will help me improve my visual storytelling and facilitation. I will be looking at further courses in Bikablo. Really recommend Work Visible - training
Solar Panel and Battery installation #
We had solar panels and a battery installed at home, so expect some boring energy posts over the next few months. Much like the panels, no doublt I will be less active and vocal when the suns stops shining! The main driver was to have more control over our energy and where it comes from. Really happy with the results so far, but the weather has been fantastic so I know I've been spoiled so far. Perhaps will look into a heat pump at some point.
AI and side projects #
I've slowed down on my AI rampage and side projects. This is mainly because I've been so busy at home and work, that it's taken all my headspace.
Over the last couple of years I've fully embraced AI tools and tutorials to learn as much as possible in a short time. From the basics of Python, Next.js, prompt engineering, AI, LLM theory, conversational design, and many AI UI tools.
There's so much to continue to learn, but as great as the emerging tech is, I do question the ultimate value of many of them. Over the last 6-12 months, I've had premium subscriptions for the following:
- Chat GPT
- Midjourney
- v0.dev
- Cursor AI
Ultimately, products and services are only as good as the idea and the user needs they are trying to solve. This has some parallels with learning I did many moons ago, after years of designing websites and apps using photoshop, then relearning fireworks, sketch, marvel and figma. So many transferable skills, and by focussing on the user needs allows us to seperste the solution from the tool required.
Picking up the camera #
With some leave planned, and a drive to create in the face of the AI wave, I'm looking firward to pick up my camera after tha hiatus of the kast few years. Hoping to do some more street photography and perhaps a new photography project.
Have a great Easter break all.
A bit about me: I'm a UCD specialist focused on Service Design, with a passion for accessibility and creating services for all. I navigate the world with a dry sense of humor, finding the funny side in the most mundane tasks.