Stacks
On Sunday morning I woke with an idea I had to try out : Plugin Universe. Basically an online database of DAW plugins. I wrote a handful of docs with the general plan.
I decided to use the core of the bloated Semem, the SPARQL and Semantic Search (embedding) infrastructure. Most of Sunday and Monday were spent ploughing through this with Claude. But then I ran out of tokens: reset on Wednesday.
So last night I went for beers.
Today, Plan B : try OpenCode. It took a little while, but it's now running with model "Muse Spark 1.3 Free". The only thing is, with this being a new setup I didn't want to mess up the already working code for Plugin Universe. For a while now I've wanted to do something useful with a list of 5,000+ bookmarks I've accumulated (on Workflowy, it's convenient to Share to..).
This time to really keep streamlined, I asked the assistant to build a knowledgebase using the core code of Plugin Universe.
OpenCode seems to be working pretty well with this model, just a handful of rate limit exceeded messages.
Heh, nice message : "...measured nonsense scores 0.562 on this corpus..."