Who went first
Tim Ballard did. Sixty days before the group posts began, Ballard posted a video and credited Latter Day Chad. This page corrects what we said on October 4 and lays out how the group around them is built.
Correction to our October 4 post
On October 4 we wrote that February 18, 2026 was “the first time Tim Ballard and Latter Day Chad appear on a post together,” and that none of Chad’s earlier posts tied to Ballard. That was wrong. On December 20, 2025, Ballard posted a video and credited Chad as a collaborator. February 18 is still the first of the seven group posts.
The claim appeared on our card “The day it started,” our Facebook post, and our first write-up, “How the Instagram apostasy started.”
The post
The header reads “timballard89 and latterdaychad.” The caption is Ballard’s own, and opens: “Brandon, are you saying my doctrine is wrong?” Instagram dates it December 20, 2025. Checked on Instagram
Timeline
| Date | What happened | Who posted | Status |
|---|---|---|---|
| Mar 2024 → | Chad and C start co-posting. By Feb 18, 2026 they share 434 posts. | — | Our data |
| Nov 9, 2025 | Ballard credits F on a post. His first post with anyone in the group. | Ballard | Our data |
| Dec 20, 2025 | Ballard credits Chad. Their first shared post. | Ballard | Checked |
| Feb 18, 2026 | First group post: five credited accounts plus the host. Ballard and Chad are two of the five. | latterdayavenger | Checked |
| Feb 20 – Apr 18 | Six more group posts. Ballard and Chad on every one; the other three seats rotate through 11 accounts. | latterdayavenger | Our data |
| Jun 22, Sep 21 | Joint posts, now uploaded by Chad himself. | Chad | Our data |
Checked seen on Instagram itself, screenshot kept. Our data from our captures of Instagram’s post data, not re-checked by hand.
The group was already there
These are the same 13 accounts as on our card “It started as a group”: Ballard (TB), Chad (LDC), and the 11 accounts that filled the other seats (A–K). Each line is a pair of accounts credited on the same post, counted across everything we have captured for them. The left map is everything before Ballard’s first group post. The right map is from February 18 on.
Before Feb 18, 2026
Feb 18, 2026 onward
What the structure shows:
- One host, assigned seats. All seven group posts were uploaded by latterdayavenger, which is not one of the 13. Ballard and Chad hold two fixed seats; the other three rotate.
- The host is a regular. latterdayavenger is also credited on 42 other posts in Chad’s and C’s grids, May 31, 2025 to June 14, 2026, usually as one of three to five credited accounts. Posts with four or more credited accounts are common across this whole crowd: 808 in our captures, 20 to 60 a month since January 2025. Ballard was slotted into a format that was already running. Our data We have not yet captured latterdayavenger’s own grid.
- An old core. Before Ballard arrived there were already 25 links and 634 shared posts. Chad and C alone had 434. Chad–D, C–D and Chad–H were also established.
- Thin new links. From Feb 18 the number of links nearly doubles, but most new ones are a single post (21 of 48). Accounts appear once or twice, then drop out. Part of that is arithmetic: one post with five credited accounts makes ten pairs at once (see the algebra below).
- Ballard hangs on Chad. Ballard has 13 shared posts with Chad and between 1 and 6 with anyone else. Before Feb 18 he had exactly two posts with this group, and he posted both himself.
The data shows who was credited with whom, and when. It does not show who is behind the smaller accounts.
The seven posts as algebra
Not yet peer reviewed
The math in this section has not been reviewed by anyone outside our team. The counts come straight from the posts; the design-theory reading and the symmetry calculation are our own and may contain errors. Corrections are welcome. The matrix and the pair counts are on GitHub.
Treat the accounts as points and the seven group posts as blocks: 13 points, 7 blocks, 5 points in each block. That is an incidence structure, and its algebra says a lot about how the posts were put together.
The incidence matrix N. One row per account, one column per post; a filled cell means the account is credited on that post. Rows are grouped by symmetry (explained below). The right-hand column is r, the number of posts each account is on.
Pairs come in bundles of ten. The pair counts are N·Nᵀ = ΣB 𝟙B𝟙Bᵀ, a sum over the posts. Each post with five credited accounts adds a complete set of ten pairs at once. So the lines in the maps above arrive ten at a time: that is the arithmetic of the format, not ten separate choices.
Deliberately unbalanced. In a balanced design every account would be on the same number of posts and every pair would meet equally often. Here r = 7 for Ballard and Chad and 1 to 4 for everyone else; the pair TB–LDC meets on all 7 posts, and only 40 of the 78 possible pairs ever meet at all.
Take out the fixed pair. Ballard and Chad are in every block, so remove them and look at what is left: seven triples on the other eleven accounts.
The first four triples all sit inside {A, B, C, D, E}, and A is in all four: a core. Then it branches, with B, D and G carrying over as bridges. HIJ stands on its own and shares no account with any other post.
Symmetry. The automorphism group Aut(N) is every way of relabelling the 13 accounts that carries the set of seven posts onto itself. Computed on these posts: |Aut(N)| = 12 ≅ S₂ × S₃, with orbits {TB, LDC} and {H, I, J}, and every other account fixed.
The order of the posts has no symmetry. The seven posts also come in time order, Feb 18 to Apr 18, and the letters A–K follow it. A time order is a chain, and a finite chain has only one order-preserving relabelling: leave everything where it is. Charles Holland studied the group of such relabellings of an ordered set (Michigan Math. J. 10, 1963). Here every one of the 12 symmetries sends each post to itself (TB and LDC are on all seven; H, I and J only on Apr 10), so they all leave the time order alone: the symmetry is in the accounts, not in when they posted.
- {TB, LDC}: from the group posts alone, swapping Ballard and Chad changes nothing. The format treats them as one unit, the two fixed seats. What tells them apart lies outside these posts: Ballard’s own December 20 post, and Chad’s 434 posts with C.
- {H, I, J}: interchangeable. They appear once, together, as a whole block: a batch swapped in as a unit.
- Fixed points: A, B, C, D, E, F, G and K each have a footprint no other account shares.
Large orbits mark interchangeable seat-fillers; fixed points mark accounts the structure depends on. This describes the posting pattern only, not who is behind any account. With latterdayavenger’s own grid captured, every post it hosts becomes another block, and the same calculation can be run on the whole design.
Check it yourself
Confirmed in SageMath 10.9 · Oct 5, 2026The seven posts and the matrix N are published. Standard, free math software can compute the symmetry group on its own, without anything of ours:
B = [["TB","LDC","A","B","C"], ["TB","LDC","A","C","D"], ["TB","LDC","A","E","D"],
["TB","LDC","A","B","D"], ["TB","LDC","F","B","G"], ["TB","LDC","H","I","J"],
["TB","LDC","K","G","D"]]
D = IncidenceStructure(B)
G = D.automorphism_group()
G.order(), G.orbits()
designs.IncidenceStructure, which Sage 10.9 does not have. Corrected October 5, 2026. Software agreeing is not peer review: what the symmetry means is still our reading.verify_automorphisms.py, plain Python with no packages: python3 verify_automorphisms.py incidence_matrix.csv prints the same 12 and the same orbits. Its README explains how the count works. (The λ on this page is design theory's pair count, not lambda calculus.)What the joint videos say
We transcribed the seven joint posts that have speech (two others are photos, one video is silent). After removing common words, these are the words said most, across all seven together:
Small numbers from short videos: scripture and gathering language (God, Zion, gathered, covenants, scriptures, Joseph Smith, Utah), plus a Korean War segment (Chinese, soldiers, Gapyeong). The names “Ballard,” “Tim” and “Chad” are not said in any of the seven.
How we missed it
Instagram’s collaborator list on a post names the credited accounts but leaves out the account that posted it. On October 4 we counted Chad’s posts whose collaborator list named Ballard. A post Ballard made himself never names him, so the December 20 post didn’t count. It was in our captures of both accounts the whole time.
The fixed rule
A post ties two accounts if either one credits the other, or if the same post appears in both accounts’ grids. Recounted that way, Chad’s 1,463 captured posts before February 18 include exactly one tied to Ballard: December 20, 2025.
Limits: Chad’s posts come from two captures merged (October 1 and October 4, 2026), 1,888 posts reaching back to June 20, 2024. Ballard’s capture reaches back to September 11, 2025. The October 1 capture did not show Ballard’s credit even on the February group posts, likely because Ballard blocks the account we viewed it from, so it cannot fully rule out an earlier tie in Chad’s oldest posts (June 2024 to January 2025). Nothing before June 2024 is covered.