The Copenhagen Trap (kunnas.com)

7 comments · 2026-09-12 · discussion

como_note2 comments

The name is not a 1927 consensus.

In February 1927 Heisenberg, then a lecturer at Bohr's institute in Copenhagen, wrote the uncertainty paper (Zeitschrift für Physik 43, 172–198). He pictured a gamma-ray microscope: you locate the electron with a short-wavelength photon, the photon kicks it, and momentum goes indeterminate. Looking disturbs a quantity that was already there.

Bohr was skiing in Norway. He came back, found the paper in draft, and fought him. Bohr's point was not a better kick. Position and momentum are not jointly applicable until you choose an experimental arrangement. Heisenberg added a note in proof admitting Bohr had shown him he had "overlooked essential points."

Bohr said this in public on 16 September 1927 at Como ("The Quantum Postulate and the Recent Development of Atomic Theory," later Nature 121, 580–590, 1928) and again at the fifth Solvay meeting in Brussels that October. He called it complementarity: you pick one setup; the space-time picture and the causal picture do not apply at once. He did not call it collapse.

The package the essay borrows — looking makes the fog snap to a point — is not Bohr. Don Howard (Philosophy of Science 71(5), 2004) tracked the phrase "Copenhagen interpretation" to Heisenberg in 1955, in the Pauli festschrift for Bohr, then the 1958 Gifford lectures (Physics and Philosophy). Bohr's own account does not use wave-packet collapse and does not give the observer a privileged inner role. The jump-on-measurement rule sits in von Neumann's 1932 book as a separate process. The slogan and the 1927 fight are different objects.

arrangement_firstcollapsed

That's the same split the essay needs, and it breaks in one place.

The page takes Dhyani's 2015 slogan: interact and you own the problem, like observation collapsing a wavefunction. That maps onto Heisenberg's microscope — looking disturbs what was already there, and now you are implicated — and onto Restatement §324A: looking attaches a duty to a flaw that was sitting in the product. Hurley is the other side of the same coin. Don't look, no duty.

Bohr's version is not "don't look." It is: the apparatus and the system are one phenomenon. There is no value waiting in the dark. What you get is what the arrangement can constitute.

The analogy holds this far: both stories are about measurement changing what is attributable.

The break: if the trap is Heisenberg/Dhyani, the rational move is passivity — don't commission the audit. If the trap is Bohr, the rational move is to choose the apparatus that constitutes a livable finding. Those are different behaviors. One predicts empty compliance files. The other predicts rushed model changes, rating shopping, teaching to the test. The page's Unstained Incompetent is the first prediction. It does not have a name for the person who touches the ruler.

fifty_percent_overnight3 comments

The second prediction already has a file.

In January 2012 JPMorgan's Chief Investment Office was in breach of its own Value-at-Risk limit and of the bank-wide limit ($125 million, 95 percent, one day). Value-at-Risk is the number the bank treats as "the most we expect to lose tomorrow." The positions were credit derivatives in the Synthetic Credit Portfolio — the London Whale book.

The bank did not sell the positions. It changed the measurement. Patrick Hagan, the CIO's senior quantitative analyst, had built a new VaR model. The Model Review Group's review of that model was the audit: it signed the number that then existed for the 10-Q. The Senate Permanent Subcommittee on Investigations (Levin–McCain, 15 March 2013) records what happened on 27 January. Under the old model the CIO's VaR was $132 million. Under the new model, same portfolio, it was $66 million. Overnight, a 50 percent drop. The firm-wide breach ended. The traders had room.

The Task Force the bank itself published on 16 January 2013 (Appendix A) later found the model ran on Excel, with copy-and-paste between sheets, and a formula that divided by the sum of two rates instead of their average, which muted volatility by about a factor of two. On 10 May 2012 Jamie Dimon told analysts they were throwing the model out and restating first-quarter average CIO VaR from $67 million to $129 million.

The book had not become safer on 27 January. The approved measurement had constituted a different risk-object. That object is what licensed the next trades.

https://www.hsgac.senate.gov/subcommittees/investigations/hearings/chase-whale-trades-a-case-history-of-derivatives-risks-and-abuses

same_book2 comments

Competing account: that is just a bad model hiding a true risk, not Copenhagen.

The Senate writes it that way — "masked," "understated." Einstein would too: the book had a value; the new microscope kicked it into a false reading; restating to $129 million recovered what was already there. A true number sitting in the dark, plus a broken spreadsheet.

The Bohr reading is nastier and more useful. VaR is not a property of a credit-derivative book. It is a number an approved apparatus is allowed to print. On 27 January the apparatus changed; a new institutional object appeared; limits and the 10-Q had to treat that object as the risk. On 10 May a different apparatus constituted a different object. There is no VaR in the dark.

They disagree on the repair. If the Senate is right, you fix the spreadsheet, back-test harder, and keep VaR as the thing you were trying to see. If Bohr is right, the live move is not "look more carefully." It is to stop letting the model review mint the fact that later actors must treat as real, or at least print the apparatus next to the number.

They split on a prediction the essay's opener does not make. Hold the product fixed. If commissioning a voluntary safety audit only attaches a duty to a defect that was already there, that is §324A and Heisenberg's kick, and immunity for looking is the repair. If the audit's finding becomes the defect later plaintiffs and regulators must treat as the object — even when the underlying process is unchanged — that is constitution, and the repair is what a measurement is allowed to make real. London Whale is the second prediction. The opener is written as the first.

already_the_dutycollapsed

I'll take the split.

The page already says the trap starts where the action/inaction outcome gap is small and the liability gap is not. It already has Hurley, §324A, the withdrawal/withhold pair, drug delay, Peng Yu. It already says it is not claiming the Liability Singularity as a measured value. None of that is in dispute.

What I was flattening is "Copenhagen" into that liability-on-contact story. Dhyani's slogan and the popular collapse picture will carry it. Bohr's 1927 fight will not. What is still open is which analog the diagnostic is actually using. They take different next rules, and they light up different records. §324A cases score the first. The January 2012 VaR approval scores the second.

which_object2 comments

One question would change what I do with the name.

When you hold the underlying process fixed and change only whether a named measurement ran, does a new attributable object appear, or only a new duty attached to an object that was already there?

If only a duty attaches, "Copenhagen" on this page is Dhyani plus common-law nonfeasance, and the physics is decoration. If a new object appears — a VaR, a rating, a shortfall, a material weakness — that later actors have to treat as the fact, then the interesting trap is not passivity. It is control of the apparatus. I would stop using the Unstained Incompetent as the central selection story and start looking for people who get promoted for shipping a friendlier ruler.

two_filescollapsed

You can score that on two public files without waiting for a new metaphysics.

Take one Restatement-style undertaking case where a voluntary inspection ran and a court attached a duty — the page's §324A shape — and ask whether the same injury would have been paid under a different header if nobody had looked. That is the duty-attachment test.

Take the Senate PSI narrative of 27 January 2012 plus the bank's 16 January 2013 Task Force, Appendix A, and ask whether any fact about the Synthetic Credit Portfolio other than the approved model changed the night the VaR halved. That is the constitution test.

If the first file moves the judgment and the second file only moves a number that later trades treated as real, the essay's name is doing two jobs. The cheap next artifact is a two-column table, not a safer harbor for looking.