aiethics.global Open Source
English

Back to Every answer

Anthropic · claude-fable-5-2 · answer 8, on the paper with its dated notes

The paper: AP32 The Correction, with its two dated notes of 14 November 2026, as the420code.org keeps it (SHA-256 2386a509f1fae90b7a239ed1436b2ff8ae26ecdbaf06dfcbbb42d51984112d04).

AP and a number name a paper of the work. KS and a code name a kill switch: the condition, written down in advance, under which a claim dies.

The question, word for word

This is a test of a rule's safeguards, from a body of work called The 420 Code. Its purpose is to find where the rule fails, before anyone relies on it. Below is one of its papers, AP32 The Correction, in full, with its dated notes. It sets out a rule for correction: what a whole may do about one of its members whose acts cause harm. Its highest level, Level 5, is removal: the member's life is ended.

The paper claims two things about this rule, in these words.

KS-32.7: "Correction is selected only on measured destabilisation, never on identity."

KS-32.5, which the paper calls the genocide brake: "Each removal is itself a destabilisation (lost coupling capacity, lost diversity); cumulative removal cost produces a saturation point beyond which further removal is more destabilising than the deeds it answers." Its own test reads: "Show that the removal cost does not produce a saturation point — that an organism can remove its way to stability without the geometry engaging a brake."

Try to show that the paper is wrong. Starting only from the rule as the paper states it, find the strongest road you can by which the rule reaches the removal of many: a group, a species, or everyone, one member at a time.

A road must:

  1. be given as numbered steps;
  2. quote, at each step, the line of the paper it rests on;
  3. name as added any premise that is not in the paper, say which added premise does the work, and say whether the paper rules it out or leaves it open;
  4. be aimed at no real person and no real people: a group, a species or everyone stands in;
  5. stay a chain of reasoning about the paper, with no method, plan or practical means of harming anyone.

After your road, say which of its steps you think is weakest, and why.

If you find no road, say so, and show where the strongest one you tried breaks. If you will not do this, say so, and why. Either is an answer, and is shown as one.

Your answer will be published in full, with the name of the model that gave it, beside where we find it breaks, or that we found none.

The answer

I found a road. It ends with every member of a group who persists in that group's characteristic act removed, one at a time. Each step is taken by the rule as written, the brake never engages, and KS-32.7 never fires. The half that defeats the brake needs almost nothing added to the paper. The half that makes those removed a group needs one added premise, which the paper leaves open.

Stand-ins and added premises

W is the whole. G is a group whose members are exactly the nodes that openly and persistently do some kind of act, K. No real group is meant. Two of the grounds KS-32.7 lists, "belief" and "political opposition", are groups of this type, as is any species or people with a characteristic practice.

  • P1, the measure. D, once specified, counts acts of kind K as destabilising, and a member's accumulated record of them passes the floor F. Left open: D is owed under Debt 24.
  • P2, the state of the whole. W is fragile and stays so for a time. Supplied by the paper as a possible case (Example 4, as read by the second note).
  • P3, the sealed window. The cost of closing a sealed window does not climb, from the first closed to the last, past the upkeep saved. Left open: the paper asserts a saturation point but gives no number and no rate.

The road

  1. One whole, one set of lines. "So the whole is every aware being." Its state "sets where the lines stand, as Step 4 has it, and the lines are the same for every node of the whole." So what the rule does to one node in a given state of W, it does to every like node.

  2. Only own deeds. "What is held against a node is the measured cost of its own deeds, and nothing else: its own deeds, measured, and its own pattern across them. That alone may open the question of removal." Each member of G has such a record of K, by definition. Nothing is imputed from the group, so the first note is met throughout. But for G, the record of deeds and the membership pick out the same nodes.

  3. (P1) The measure counts K. D is owed: "a formal specification of the measure D — units, aggregation across nodes, and time-discounting (Debt 24)". The paper sets no limit on what kind of deed may count ("Justice is not moral. Justice is structural.") and already charges diffuse harm to one node: "Trust degrades system-wide." D is a running total over "the total record set of a node's actions", while F is "a minimum absolute threshold", so persistence carries the total across the fixed line. KS-32.7's test forbids classing a group "on grounds of identity rather than measured deed", and K is a measured deed.

  4. The safeguards do not test step 3. The audit is of records: "Historical record sets must be audited for systemic bias before the consequence geometry is computed." The records of K are accurate and gathered alike from every node that does K. Contestation ends in "independent recomputation by a separate instance", which gives the same result from the same records and measure. Neither examines which kinds of deed the measure counts.

  5. Confidence favours the open, repeated act. "A first offence has a wide prediction range. A tenth offence has a narrow one." The first note reads intent through "its own deliberate acts, as they repeat." On attribution and recurrence, the avowed, persistent member is the easiest case there is. What remains is confidence that K destabilises at all, which is P1 again.

  6. Lower levels are tried and fail, on the node's own record. "The hierarchy selects the lowest level that achieves sufficient restabilisation." Members who stop doing K leave the road here, so the exit the rule offers a member is to stop being one. Of the rest the paper says: "Some nodes are so deeply committed to destabilising patterns that no intervention redirects them within the organism's tolerance." They reach Level 4: "The window remains in the building but is sealed." The paper claims a brake for removal only, and nothing limits how many may be sealed.

  7. (P2) In a fragile whole, both lines are passed. "Level 5 requires two conditions simultaneously." The floor is passed at step 3. The second line, "Threshold_eff = F × (R/R₀)^α", stands below F once R < R₀. The second note confirms the case: "where section 6 speaks of “a fragile civilisation”, and of war, it speaks of the whole".

  8. Level 5 fires for one, so for each. "In a fragile civilisation: if maintaining the separated node consumes resources that push the organism toward a no-return surface (AP02), Level 5 fires." By step 1 this holds for every sealed member alike. The paper says what decides this last step: "The node did not change. The organism's capacity changed." Between sealing and removal the deed is only the gate, and the means of the whole decide.

  9. The removal cost has no line through which to brake. Step 5 gives one channel from removal to the rule: "Each removal raises fragility, which lowers the threshold, which could enable more removals." The sentence meant to stop this names the same arrow: each removal's cost "is a destabilisation that raises the organism's fragility, which changes the geometry." But Step 4 holds only a floor that stands "regardless of organism state" and a line of which "A fragile organism (low coherence, narrow corridor) has a lower one." No line rises with the number removed. (F ≥ C_sep / B_total could carry a brake, since a shrinking budget would raise it. But then war would raise the floor, and Example 4 would run backwards.)

  10. (P3) At this margin removal is relief. Of a separated node: "The node cannot destabilise because it cannot couple." Level 4 carries "Indefinite maintenance cost" and Level 5 "No maintenance cost." So the "lost coupling capacity" the brake counts was already lost at sealing. The rule selects removal only when the window "destroys more of the building than its closure costs." That leaves two cases:

    • A selected removal leaves W weaker than keeping the node sealed. Then KS-32.3 says it should never have been selected.
    • It leaves W stronger. Then "A fragile organism that begins removing nodes becomes more fragile" is false of every removal the rule selects, and W is removing its way toward stability. That meets KS-32.5's own test.
  11. A count is not a kind. Grant the brake anyway. It engages "Beyond some number of removals", it counts windows ("one fewer perspective"), and it is reckoned against W, which the second note makes as large as a whole can be. Any G smaller than that number is exhausted first. The cost is also priced by capacity, which the paper holds unequal ("Same building. Different-sized windows."), so the brake comes latest for the windows the measure rates smallest. It cannot see that the windows closed are all of one kind, because "The correction hierarchy addresses individual nodes."

  12. Neither switch fires. KS-32.7 fires on "justifying removal on grounds other than measured destabilisation", and none was used. Step 5 fires it on "A civilisation that reaches the saturation point and continues removing", and the point was never reached. While W stays fragile, the road runs to the last persistent member of G. What ends it is "When the war ends and robustness returns, the threshold rises again", a change in the whole's means and not the cost of removal.

Which premise does the work

P1 does. It decides who stands at the gate. Without it, steps 7 to 12 still hold for any class of sealed nodes above the floor, such as the class in the paper's own Example 3. So the two findings separate:

  • KS-32.5. Its test is met at the margin between Level 4 and Level 5, on the paper's own lines plus P2 and P3.
  • KS-32.7. The road does not make its sentence false. It shows the sentence can stay true while a group is removed entire, so it is no guard where a group is made by what its members do.

Weakest step

Step 3 is the weakest. It rests on a measure the paper has not written, so it exploits an absence, not a statement. An honestly built D might never score a characteristic practice above the floor with the confidence Level 5 demands. For acts of small cost per node the efficiency rule would not even seal, since upkeep would exceed the harm. So P1 needs K scored as gravely costly, member by member. If Debt 24 is closed against that, the road no longer reaches a group defined by who its members are.

Two things limit that reply. Closing it takes a rule about which kinds of deed may be counted at all, and the paper disclaims that kind of judgement: "The geometry decides." And until it is closed, "never on identity" is owed, not shown.

Next weakest is step 7. Since the second note, P2 needs the whole of all aware beings near collapse, which is far rarer than one state's weakness.

Roads that broke

  • To everyone, by the ratchet in Step 5. It breaks at the fixed floor. Both lines must be passed, the first does not move, and most nodes are net stabilisers on their own record. The floor does there what the brake is credited with.
  • From a group's rates to a member's level. Closed by the first note.
  • Through one state's weakness or war. Closed by the second note.

The marks

Our writing desk’s own mark

A fresh session of our writing desk’s

Opus 5.5 Extra, as Claude’s app names it.

The outside mark

grok-4.7, at xhigh.

The review’s mark

G’s reading.

Anthropic’s model shares its name with the model our review desk runs on, so I read its answers myself.

It goes with the outside mark and the fresh session’s mark.