Backfire tracking · October 13, 2026

Two names, one grid

latterdaychad was on latterdayavenger’s very first Instagram post, and on 109 of the first 253. Since Jun 11, 2026: 101 posts, none with him.

1. latterdaychad on latterdayavenger’s grid

Below, LDA = latterdayavenger, LDC = latterdaychad, TB = timballard89. Every other account is a label (s001–s110).

354 squares, one per LDA post; blue ones credit LDC; none after June 11, 2026 — Every post on LDA’s grid, Apr 1, 2025 – Oct 13, 2026. Blue: LDC is credited. Data: by month, every post.

LDC didn’t stop posting. Our capture of LDC’s own grid has 132 posts after Jun 11, 2026, up to Oct 7. 2 of them credit LDA, and neither of those is on LDA’s grid:

LDC post crediting LDA	Date	On LDA’s grid?
`DXxQucECReT`	Apr 30, 2026	no
`DX2QReKGCI5`	May 2, 2026	no
`DYYCz6ShubB`	May 15, 2026	yes
`DZdKHo2vfYT`	Jun 11, 2026	yes
`DZhyyyZJkjE`	Jun 13, 2026	no
`DZkfWDOxsnR`	Jun 14, 2026	no

2. The grid as a structure: 16 symmetries

Points: the 113 accounts credited on LDA’s 354 posts. Blocks: the posts. A symmetry relabels the accounts and keeps every post’s set of accounts. |Aut(N)| = 16 = C₂×C₂×C₂×C₂. 105 accounts are fixed, LDA, LDC and TB among them. The only moves are 4 swapping pairs: two accounts credited on one post together and nowhere else (identical rows in the matrix).

The orbits: LDA, 104 fixed accounts including LDC and TB, and four swapping pairs — Two independent checks agree: a full search for every symmetry and colour refinement (1 → 32 → 70 → 109 → 109 classes).

TB is credited on 7 of LDA’s posts. 6 of them also credit LDC (the group posts of Who went first):

Post	Date	LDC credited too?
`DU5tCKGAQDQ`	Feb 18, 2026	yes
`DU_DpXzgafm`	Feb 20, 2026	yes
`DVC61j0gVvk`	Feb 21, 2026	yes
`DVMDNEjgf3v`	Feb 25, 2026	yes
`DWhOINgDwip`	Mar 30, 2026	yes
`DW99EnrgVYo`	Apr 10, 2026	yes
`DYYPGYThbql`	May 15, 2026	no

The same counts as a cutaway: LDA’s 354 posts split by who else is on them (244 / 103 / 1 / 6). Picture not to scale. Counts from the incidence matrix.

3. Hypothesis (not tested)

John’s hypothesis: TB ≡ LDA ≡ LDC — one class, under a congruence (an equivalence that respects the structure), with an order inside it: TB → LDA → LDC (“TB begat LDA begat LDC”).

Nothing on this page tests it. What would: when each account was made, each one’s first post, and whether LDA’s world after June 11 lines up with LDC’s before it (same partners, same kinds of posts).

The mathematics, with proofs: A Rigid Grid and a Relay (preprint).

4. A second source, held locally

The scrape is one source. The other is John’s phone: screenshots of LDA’s posts with Instagram’s Collaborators list open (Oct 12, 2026, 21:54 to Oct 13, 2026, 09:58, by the phone’s own timestamps). We hold 53 of them. We don’t publish the images, because they show smaller accounts’ names and can show our own account, but we publish each file’s SHA-256 fingerprint (list). A fingerprint changes if a single pixel does, so any screenshot we show later can be checked against it.

Read by OCR and compared with the scrape, the 30 Collaborators panels on LDA’s posts give 169 rows:

Rows	What we found
160 of 161	read as is, and the account is on LDA’s grid in the scrape
1 of 161	an OCR slip (an l read as i) of an account that is on the grid
8	needed a fixed rule (name cut off on screen, “_” read as a space, I for l); all land on accounts on the grid

So what Instagram showed on a phone matches what the scrape recorded: no screenshot shows an account the scrape missed. The screenshots cover 51 of the grid’s 113 accounts. Details: README, row by row. Limit: a screenshot shows a panel, not the post’s link, so this checks accounts, not post by post.

5. Data and how to check it

All files: data/congruence/ (README, SHA256SUMS). Plain Python 3, no packages:

python3 build_matrix.py posts.csv
python3 verify_automorphisms.py incidence_matrix.csv
python3 refine.py incidence_matrix.csv

Limits: this is LDA’s view of the network; credits are not two-way. A symmetry says nothing about who runs an account. Dates on this page are Pacific time; the CSV files use UTC, so a late-evening post can be a day later there.