A lot has changed at Scootercam in the last few weeks – almost everything, really. Here’s the rundown:
New hosting provider
We’ve moved from Hostinger to GlowHost and despite some rough first days, it’s looking good. GlowHost gives us access to ffmpeg, a utility that assembles our timelapse segments. Hostinger provided ffmpeg as well, but also required a ton of overhead work – work that GlowHost provides as part of the package. GlowHost, from that perspective, gives us the same needed service without making us fly blind through server management.
New content management
Throughout Scootercam’s almost 10-year run, it’s been a sandbox – I could code, and learn and evolve the site along with my skills. But PHP hobby-code Is not manageable by anyone other than the hobbyist. If I wanted Scootercam to have any chance of long-term viability, I’d have to separate the code that does the work (assemble timelapse videos, or produce a forecast) from the code that builds the pages. By adopting WordPress as my architecture, others could conceivably run the site – as long as ‘the code that does the work’ works as dependable WordPress plugins.
AI guidance
Which brings us to AI, and vibe-coding. A few months ago I asked claude.ai to create and refine some code for me – code that took a text-based forecast from the National Weather Service and broke it down in to parts I could use. My script was rough and I was pretty sure there were nuances about the incoming file I should, but didn’t know. I asked Claude to give it a shot, and the result was much better, much more efficient than my original.
Before long I was ‘vibe-coding.’ I was asking Claude to create scripts not by improving or analyzing existing code, but by responding to my natural-language descriptions of what I wanted it to do. It’s hit-or-miss, mostly, and does go its own way sometimes. But Claude has been successful enough to stay on the team for now, and has been tasked with writing the plug-ins and README files that the new Scootercam WordPress site uses.
Upcoming
I have ordered a new weather station so I can decommission the current one-off system of Raspberry Pi-based sensors. I’ll describe that once we’re online with it, hopefully before Thanksgiving.