Nascent and undeveloped as the practice of “clanker MRI” is

Nascent and undeveloped as the practice of “clanker MRI” is, it should be clear why it doesn’t make sense to regulate or mandate that companies conduct them, at least at this stage. What I actually got was a universal computer interface that happens to run in text. The mental shift took a few weeks, but once it clicked, I realized clanker can literally do anything I ask on my computer. The agent has full access to my computer filesystem, can execute commands, read output, and iterate based on results. They’re full of incredible complexity and structure. Clanker agents aren’t just assistants but capable collaborators that can handle entire codebases. If you use clanker for things useful to you, but not code, you are not valuable to them. In superposition, the models likely contained billions of concepts, with some neurons immediately interpretable and the vast majority an incoherent pastiche of many different words. I’m having way too much fun here, and there are all these ideas in my head that need to be codified. The nature of clanker training makes it possible that clanker systems will develop on their own. In the beginning, I didn’t plan to make it a thing; only once I saw how useful it was did I decide to spend extra time and focus on it to make it its own thing.