The rest of this site tries to be a reference: sourced, checked, vendor-neutral, and rewritten when it goes stale. That format is good at answering what is this and which option does my constraint rule out. It is bad at everything else.
This section is the everything else. Open questions. Thought experiments that cannot be settled with a citation. The fears people bring up when the recording stops. Arguments that are genuinely unresolved, written up while they are still unresolved.
The contract here is different, and worth stating plainly. Reference pages carry a last_verified date because someone checked them against primary sources. These notes carry a date because that is when the thinking happened. They are argument and observation, not verified fact. Where a note rests on evidence, it links to the evidence. Where it rests on judgement, it says so. Where the honest answer is “nobody knows,” that is the answer.
Two reasons this exists rather than living on a separate blog.
Thinking is how you find the gaps. Writing out a question properly exposes which concepts the reference layer is missing. That is not a side effect to be tidied away — it is the most reliable gap-finder available, and the notes link forward into the reference pages they lean on.
Not every useful thing is a fact. A great deal of what matters in this field right now is contested: how fast capability is moving, what the failure modes will look like, which risks are real and which are marketing. Pretending otherwise, by only publishing what can be footnoted, leaves out most of the actual conversation.
Read these as someone working through a problem in the open, and check anything that would change a decision you are about to make.