1
00:00:00,200 --> 00:00:02,801
Matt: Compitum routes under hard resource limits.

2
00:00:03,150 --> 00:00:06,203
Matt: It had shadow prices, but did not let them choose.

3
00:00:07,100 --> 00:00:11,280
Ellie: FabricPC asked: can hidden scarcity become a routing price?

4
00:00:11,650 --> 00:00:14,738
Ellie: The proposed chain ended with price choosing action.

5
00:00:15,150 --> 00:00:16,613
Matt: But the final link broke.

6
00:00:17,000 --> 00:00:22,190
Matt: Even exact belief failed when one local price was multiplied by a lumpy action.

7
00:00:22,550 --> 00:00:27,697
Matt: The predictor was not the whole problem. The steering rule itself was wrong.

8
00:00:28,200 --> 00:00:30,557
Ellie: The correction prices the whole action.

9
00:00:30,750 --> 00:00:34,790
Ellie: It compares future value before spending with future value after spending.

10
00:00:35,000 --> 00:00:40,689
Ellie: The discrete unit charges telescope to the same two lookups: exact charge, not a tangent.

11
00:00:42,150 --> 00:00:47,618
Matt: Immediate utility minus action charge now ranks choices exactly like Bellman.

12
00:00:48,150 --> 00:00:52,542
Matt: Certified choices and utility matched. Mismatches: zero.

13
00:00:54,050 --> 00:00:55,571
Ellie: Belief still matters.

14
00:00:56,000 --> 00:01:02,513
Ellie: At one reachable state, belief point zero five selects spend; point two zero selects opportunity.

15
00:01:02,850 --> 00:01:05,961
Ellie: This proves an action change, not the threshold.

16
00:01:06,300 --> 00:01:09,748
Matt: The fixed FabricPC predictor did not clear its gate.

17
00:01:10,000 --> 00:01:16,269
Matt: Predictive coding tied backprop at point three one four: fifteen point four percent of the recoverable gap.

18
00:01:17,050 --> 00:01:20,220
Ellie: The intervention restored exact action pricing.

19
00:01:20,650 --> 00:01:23,649
Ellie: The model lost. The experiment succeeded.