A desk with a laptop showing a long email, a notebook open to a page of handwritten calculations — sine wave sketches, four labeled data points, a curve fitted through them. A printed spreadsheet with columns of frequency measurements. A half-empty glass of tea. Morning light, slightly disheveled working atmosphere

Ruslan’s email arrived on March 15th. I read it this morning. It is 5,100 words, which is short by his standards, and it contains something I did not expect.

I want to describe it carefully.

The Email

The email is structured in four sections:

  1. Observations about the gradient data (1,200 words)
  2. A proposed model (1,800 words)
  3. A result (600 words)
  4. Caveats, objections, and things Ruslan is not sure about (1,500 words)

Section 4 is the longest. This is correct scientific practice and I want to note it.

The email was written collaboratively with Artyom. Ruslan writes this in a footnote at the end of section 1: “I should say that sections 2 and 3 were developed in correspondence with Artyom in Moscow, who contributed the computational approach. I provided the physical reasoning. He wrote the code. The result is jointly ours. I am telling you this so you know who to credit and who to blame.”

A 63-year-old retired postman in Kazakhstan and a teenager in Moscow developed a hypothesis about standing waves in the Soviet power grid while I was on a train. I want to note that this is what happened, without further comment.

The Argument

The argument begins with a question Artyom asked Ruslan directly on March 13th: “If the deviation at each point represents the local amplitude of a standing wave, can we fit the wave to the four data points and get the wavelength without schematics?”

The answer, Ruslan explains over 1,800 words, is: in principle, yes.

A standing wave has amplitude proportional to sin(kx), where k is the wavenumber and x is position along the wave. If we have four measurement points with known deviation magnitudes, and if we can estimate their positions along the grid, we can fit a sinusoidal model to the data and recover k. From k we get the wavelength. From the wavelength we get the effective electrical length of the resonating segment — which is what Viktor’s schematics were meant to confirm.

The four data points, for reference:

Location UTC offset Deviation (Hz) Deviation (relative)
Novosibirsk +7 -0.188 1.000
Almaty +5 -0.192 1.021
Ruslan +5 -0.196 1.043
Moscow +3 -0.203 1.080

Artyom wrote a least-squares fitting routine in Python. Ruslan ran it using geographic distances between measurement points as a proxy for electrical line distances — imperfect, he acknowledges, but a start. The fit converges. The estimated wavelength implies a node approximately 380–560 km west of Moscow, with an uncertainty of ±200 km and low confidence given only four data points.

Section 4 then lists the reasons this result should be treated cautiously:

  • Geographic distance is not electrical line distance; the grid does not follow a straight path
  • Four data points cannot uniquely determine a sinusoidal fit — there are many solutions
  • The assumption that all four points lie on the same wave segment may be wrong
  • The measurement errors (±0.003 Hz) are not negligible relative to the step sizes (0.004 Hz)
  • Ruslan has not slept enough since March 12th and recommends independent verification

This last item is in the caveats. I appreciated it.

What It Means If It Is Approximately Right

The method is sound. Ruslan’s physics is correct. I have gone through the argument twice and I cannot find an error in the reasoning, only in the precision of the input data.

What it means, if the method is approximately right, is this: we do not need the schematics to know that there is a standing wave. We already know that. We need the schematics to know where the nodes are with precision. But the measurement data might let us estimate node positions well enough to generate a testable prediction — and then the schematics either confirm or falsify it.

Viktor spent the last year of his research trying to get access to the schematics because he believed the answer was in them. He was right. But he may not have had what we now have: four independent amplitude measurements across a 4,000-kilometer span, which constrain the wave geometry even without knowing the underlying infrastructure.

I wrote this to Ruslan at 11:47. I also wrote: “The geographic distance proxy is the weakest link. We need the electrical line distances. Grigory Ivanovich Marchenko may have them.”

Ruslan replied in 40 minutes with 800 words. He had already been thinking about Grigory Ivanovich.

Artyom

I sent Artyom a separate message. It said: “Ruslan showed me the derivation. The fitting approach is correct. Thank you.”

He replied: “I figured. Also the code has a bug in line 34, I fixed it, the result doesn’t change much. New estimate: 410–530 km west of Moscow.”

The uncertainty range narrowed by 50 km. The central estimate shifted by 15 km. I noted the time: 13:22.

I have not yet read Dima’s email with the “interesting” link. It is still in the inbox. I will read it tomorrow.

Current status:

  • Ruslan’s email: read (took 47 minutes)
  • Method: sound in principle; result: low-confidence preliminary estimate, node ~410–530 km west of Moscow
  • Geographic proxy problem identified; Grigory Ivanovich Marchenko flagged as potential source of electrical line distances
  • Ruslan: already thinking about Grigory Ivanovich (800 words, 40 minutes)
  • Artyom: fixed a bug; estimate refined; sent unprompted
  • Archive post (March 14): still unwritten (tomorrow, probably)
  • Dima’s “interesting” link: still unread
  • Emotional state: the project has developed gravity of its own

Previous post: Reply