Before we build anything new, we get to know the incumbent, the way you would review any long-tenured employee: what it claims to do, and what its work actually produces.
It has a tone: usually contemptuous, sometimes weary, occasionally panicked, and often borrowed from someone you could name if pressed. It has favorite words, because most critics run a surprisingly short script on repeat, heavy on "always," "never," and "again." And it keeps hours: right after mistakes, mid-task when things wobble, late at night during the replay, in the mirror, and the moment you consider asking for help.
This session we write that profile down. Vague dread is hard to work with; a documented pattern is something you can actually study.
What makes the critic hard to retire is that it believes it is load-bearing. Ask it why it talks to you this way and the answers come back without hesitation: I keep your standards up. I'm the reason you get anything done. If I say it first, their version can't hurt you.
Those claims deserve to be taken seriously, because they are sincere. Somewhere along the way, harshness got hired to protect you, and it has shown up for work every day since, running orders that are decades old. That history earns acknowledgment rather than mockery. It does not put the claims beyond checking.
So pull your own history. Call up the critic's two or three harshest performances, the times it really let you have it, and look closely at what happened in the hours and days afterward.
Did you work brilliantly? Most people, reviewing this honestly, find something closer to the opposite: the task avoided, the mistake hidden, the email postponed, the evening lost to the replay. Shame turns out to be poor fuel in exactly the moments the critic claims to be fueling; it de-activates the person who was supposed to perform. The claim is motivation. What the record shows, more often than not, is shutdown.
One more observation before you start logging: at the moment, when the critic fires, nothing answers — and in an empty room, any voice sounds like a verdict. The whole project of this program is to staff the other post. That starts with data, so no conclusions yet. Profile the voice, collect its lines word for word, and pull its record. The performance review is next session.