Introducing the DCSki Time Machine (Beta)
3 posts
2 users
127 views
2 users
127 views
I've just added a new beta feature to DCSki that I've been having a lot of fun exploring: the DCSki Time Machine.
I thought it would be interesting to provide a historical view of what DCSki readers — and DCSki itself — were talking about during each ski season. For purposes of the Time Machine, I define a ski season as running from May 1 through April 30, so the 2005-2006 season, for example, covers May 1, 2005 through April 30, 2006.
The Time Machine currently goes all the way back to the 1999-2000 season. Pick a season and you'll find an overview of what was happening that year, a chronological timeline of significant dates and discussions, and a collection of major themes that emerged over the course of the season. The timelines link directly into the original DCSki Forum discussions and articles, so it can be a fun way to disappear down some historical rabbit holes.
Building this required analyzing a pretty enormous amount of material. Behind the scenes, I developed a multi-stage process that combines conventional database analysis with AI. It first works through summaries of Forum discussions to identify potentially significant topics, then takes a deeper look at selected discussions, reviews every DCSki article published during the season, and finally brings those pieces together into the season overview, timeline, and themes you see in the Time Machine. The original Forum posts and articles remain the source material; AI is being used to help sift through and organize decades of it.
I'm calling this a beta because I can already think of ways to improve it, but I found the initial results interesting enough that I wanted to share them now. There will undoubtedly be some errors and omissions. There's also an important limitation to keep in mind: the Time Machine can only reflect what people chose to discuss in the DCSki Forum and what DCSki chose to publish. Some subjects received extensive coverage while others received little or none, so these summaries inevitably inherit some of those biases and editorial choices.
One thing that made this project possible is that I've tried over the years to preserve DCSki's history. Forum discussions have survived several generations of Forum software, and every DCSki article going back to the first one published in 1997 remains available on the site. Having that archive intact gives me some interesting possibilities for finding new ways to explore it.
So grab a cup of coffee, pick a season, and take the DCSki Time Machine for a spin. You'll find the option in the main Forum menu, sitting between "Lonely Topics" and "Search". Or just click here to begin exploring. And please use this thread to share your thoughts about the feature — or, even better, some of the memories that come back as you explore old seasons.
👍 1
so it begs the question - from me. What did we go with given you trained the system using database content and text? AWS (maybe Bedrock), ChatGPT API, Claude (maybe not) ?
At any rate 💖 love the redesign! Did you use like SASS/Compass to generate the CSS or did it by hand? Code looks great very clean!
👍 1
Thanks! I went with the OpenAI API, integrated directly into DCSki's PHP backend (although written in a way that I can easily switch AI models). There's no AWS/Bedrock layer involved. I've built a common AI interface on the server that I'm now using for several DCSki features. (DCSki is still running on a dedicated, colocated Mac Mini. It's been a great server!)
For the Time Machine, it's a multi-stage process rather than simply throwing a season's worth of text at a model. DCSki already has AI-generated summaries of Forum threads stored in MySQL, so I use those for a broad first pass across an entire ski season. From there, selected discussions get a deeper analysis of the underlying posts, while a separate stage analyzes every DCSki article published during that season. Those reviews produce structured data — significance, themes, notable developments, etc. — which gets stored back in MySQL.
A final AI curator pass works across all of that research and generates the season overview, timeline, major themes, and links those conclusions back to the relevant original Forum discussions and articles. The whole pipeline is checkpointed, so the various stages can take hours without losing completed work if something gets interrupted. The finished result is stored as structured JSON in MySQL; viewing the Time Machine doesn't make any API calls.
And thanks for the comment on the CSS! No SASS/Compass. It's regular CSS, written directly as part of the redesign (with some assistance from AI). DCSki is still very much a custom PHP/MySQL application rather than something built around a modern front-end framework. I've been gradually modernizing the HTML/CSS while keeping the underlying site and nearly 30 years of content intact. It's been a robust framework all these years and still gives me the flexibility I need.
I've never said that I'm not a nerd. 😂

