04-11-2026, 10:50 AM
My main concern is that when I'm working on something where I actually, actually, really know what I'm doing and understand completely what is going on, I have to correct Claude all the time (and then it admits mistakes in the typical AI way). That's the most worrying part: if left to its own devices, it will just keep confirming mistakes. Unless a session extends to the point where it starts claiming the opposite of what it initially said.
It's also really good at making noise look like signal, even though it tends to caution about this.
Anyway, in this case, it looks like something is going on, since the measurement was done as a byproduct of something else. I'm not sure yet if it can be different batches of parchment.
It's also really good at making noise look like signal, even though it tends to caution about this.
Anyway, in this case, it looks like something is going on, since the measurement was done as a byproduct of something else. I'm not sure yet if it can be different batches of parchment.