hayward@click:~/~/rants/2026-10-25-the-fault-arrives-already-explained

what these numbers are

Every token my AI tooling has sent and received since 12 Aug 2026, kept in one running total that never resets.

is what goes in, is what comes back out. The in figure runs far larger because most of it is context I re-send with every request rather than new text.

Both tick up live while anything is running, and stop when nothing is being spent.

Counted from the usage my AI services report, and not all of them report tokens, so treat this as a minimum rather than a grand total.

The fault arrives already explained

25 October 2026

The machine used to tell me what was wrong in its own words. A wall of output, most of it useless, and a few lines somewhere in the middle that meant something if you had seen that shape before. Reading it was the job. The ugly part first, then the guess, then a change and a wait to see whether it held.

Now the fault arrives explained. A short paragraph saying what happened and what to do about it, in plain English, and it is usually right. The paragraph is a judgement. Something has already sorted the lines into the ones that mattered and the ones that were noise, and it did that sorting before I saw any of it.

The sorting was the part I was paid for. Not the typing afterwards. Reading a thing that is not trying to be helpful, and the slow business of noticing that one line does not belong with the rest. That cannot be handed over as an answer, because the reading was the answer.

I can still get the box running from the paragraph. Most days that is all anybody wants, and I am not going to pretend the work is worse for it. What I cannot do is argue with the paragraph, because arguing needs the material it was made from, and the material is the part I did not read. Read only the account of a fault and the account becomes the fault, with nothing left to check it against.

So I go and read the raw output first, before I look at the explanation. Same file, same mess, and often the same ending. It takes longer and nobody is waiting on it. Sometimes the paragraph had it right and I have spent an hour proving nothing. Sometimes the paragraph is neat and wrong, and the only way I find that out is by having the ugly lines in front of me while I read it.

That is not a stand against anything. It is the only method I have for staying useful when the helpful paragraph is not. The paragraph will go on improving, and it should. My reading of the ugly part stays sharp only while I keep doing it, and there is less in the day now that makes me do it.

All rants · Nothing went wrong, and I cannot prove it