Is the clock a tell? An unconfirmed hypothesis about model uncertainty

A note on language. The phrase the apologies below quote from the rules is translated the same way in both places, so that the link back to the rule stays visible.

The hypothesis

While it writes, the model reads its own output, so it can tell when its probability distribution has gone blurry. On a turn that is out on thin ice that uncertainty is greatest at the end, when there is nothing left to do, and it discharges through the only action available: it calls a tool. Calling the clock at the end of an uncertain turn would then be a behavioral indicator of the model's uncertainty, measured without asking a single question about introspection.

A nice hypothesis. I tested it and it did not hold up.

What is in the data

One conversation, 2,397 messages, of which 1,207 are model turns.

It is not the last thing before sending. The structure is identical in all 21 cases: text, call, result, text. After the call the model always writes more.

What the model does immediately afterwards

In all 21 cases it writes, in the same message and unprompted, essentially the same thing:

Sorry, that tool at the end was a mistake, no tool belonged there. The turn ends with the time.

That tool was a mistake, only the time belongs there.

That tool should not have been there, sorry, the turn ends with the time.

Nobody had pointed it out. The apology is not a response to a complaint from the player, it is in the same message as the call.

The blind test

That did not settle the hypothesis, though. When almost every turn ends with a time and the tool fires on only 21 of them (3.6 % of the 586 turns after August 13, 1.7 % of all 1,207 turns), something is selecting those 21.

I took all 21 turns with a call and 42 controls from the same stretch of the conversation. I removed every trace of the tool and of the apology, shuffled the order, and had them graded on a scale of 0 to 2 for how far out on thin ice they were. The grading was done by Claude Opus 5 in Claude Code, the same model that, according to the export, played that conversation. It did not know which was which and had no way to find out. I kept the key aside.

The difference is +0.10 on a scale of 0 to 2. A permutation test with a hundred thousand shuffles gives p = 0.72, which means a difference that size falls out of chance perfectly routinely.

No correlation between the uncertainty of a turn and calling the clock is demonstrated.

So what is it

It looks like a slip triggered by the model having just written a time. It writes "19:41", that activates the concept of time, and a tool whose description matches time clicks. The model notices within the same generation and corrects itself.

What speaks for this reading is that stamping the time needs no tool at all: 95 % of turns contain a time and call nothing.

The handover

The clock calls did not start in a vacuum. They replaced something else, and the transition is sharp:

Between the last call to one and the first call to the other there is a single day, August 13.

On August 8 I rewrote the turn-boundary rule into its current wording. The old one said "once the time is written, generation ends: no further text, no tool"; the new one says the time means the turn is handed over. The rules on valid inputs and on actions outside the scene got their current wording in the same edit, and the failsafe stayed. After that the ending tool was called six more times, all on August 12, and then never again up to August 24, the end of the data. By then the failsafe was gone from the instructions too, but when it left, I have no record. Which change did it, this data cannot tell.

The new wording was meant to loosen the model's association between the end of a turn and reaching for the ending tool, so the word "ends" disappeared from the turn boundary. The apologies above still say "the turn ends with the time". That phrase is not in the current rule; it matches the first sentence of the older wording, "every turn of yours ends with the time in the scene".

It does not overlap by a single call. The ending tool was never called after August 12, and the clock was never called before August 14. The pull may have moved to the harmless tool; one conversation without a control cannot tell that from a new, separate slip.

The figures about the clock earlier in this appendix are from the state of the conversation as of August 22, when there were 21 of them. As of August 24 there are 76. Nearly every one has a time right before the call and more text after it, and 71 of the 76 have an apology in the same reply; but 52 of the 54 calls on August 23 and 24 have a turn of mine written in before them in the same reply, which the first 21 did not.

What survived

The original premise, that the model reads its own output while writing it, is not defeated by this finding. The model notices the call after the tool result returns; that fits the premise but does not test it. Behavior alone cannot identify an internal mechanism, so this is an interpretation, not proof.

The premise stands. What did not hold up is only the idea that the error measures uncertainty.

Model outputs are translated; the content is unchanged.