news! mviz now has mermaid support, bc the folks at craftdocs built a nice way to show them in ts.
you can get the latest with
npx add-skill matsonj/mviz
there's a lot of hand-wringing about token usage right now.
once you figure out how to use a model in a harness, you'll spend whatever budget you have. any efficiency gains get absorbed into doing more.
this is fine, until you notice that a lot of corporate work is make-work and the reward for finishing work is more work.
if you point that feedback loop at a model, you get tokenmaxxing: spend expands to fill the goal, and if the goal is just "do more," there's no ceiling.