The wrong question, for three weeks

The question since the first week of July has been what this microcode does, and for three weeks I asked it with statistics, because statistics is what you use when you cannot read. Entropy per byte position, modulo every N I could think of, across all 63,744 bytes of the master: even N splits into two populations and odd N does not, which tells me the unit is one word wide and nothing whatever about what is inside it. Then mutual information between neighboring bits, with minima that fall after bit 4 and after bit 12, and I have not believed those for days.

Tonight I looked at what I have instead of at what I wanted. The histogram of the top four bits is flat through the middle with no dominant idle encoding, and real fields have favorites, the way a kitchen drawer has three things you touch and nine you do not. So either the field is not there, or it is not where I put it. If it is not there, then the layer I have been measuring all month may not be a layer, and I do not have a better question to hand.

This is the part that will not go down. It was not too hard for me and nothing was missing: every byte was on the desk, the tools worked, and I read them wrong for three weeks and kept reading after the answers stopped making sense, because I had decided in advance what they would say. The commit went in at ten to eleven, field boundaries at 4 and 12, and I do not believe them, and then I sat there with nowhere to put any of it. It went into my body the way losing always goes into my body. I got my jeans open without standing up, put two fingers between my legs and did it fast and hard, heel braced against the edge of the desk. I was wet before I got there, which I stopped finding interesting years ago. It was not about anybody. I came without a sound, eyes open on the flat histogram, and it moved nothing except where the pressure sat.

Afterward I stood at the kitchen window in the dark and smoked, which I do seldom enough that the pack is always stale. Three weeks. Not one instruction.