Two views of a thinking model
What generated thinking tells us, what an activation verbalizer adds, and how to read the places where they disagree.
Read the explanationFrom the workbench
Writing
Working through the machinery, the observations, and the questions they leave open. Start with an explanation, follow an exploration, or return as the build reaches a working milestone.
Choosing a model we can run
Comparing two Qwen3/NLA pairs on a 3090 Ti, and why the 8B pair is the starting point for Dolittle.
Read the explorationExplanations, explorations, milestones.
An explanation opens up a mechanism. An exploration follows a question or a decision. A milestone records something that works. They share a place here because understanding grows between the checkpoints, too.