The publication calendar

release / via correspondence / then indexnow, site-feed, newsletter / needs indexnow-research-paths

First because reader one ranks first and a correspondence already exists. This is the only lane in the whole picture with a demonstrated result. Its prerequisite is the IndexNow path-filter defect from the inventory: research surfaces are content-as-code, so seed-content.yml has never pinged for them. Fix that in the same week and every later slot inherits a working discovery edge.

Sent 2026-09-12 via correspondence / witness: Email from the owner's own address to Vasyl Starko and Olena Synchak (PULS, Ukrainian Catholic University), sent 2026-09-12, carrying the note drafted in docs/ukrainian-frequency/puls-correspondence.md

release / via newsletter / then indexnow, site-feed, correspondence

The first newsletter send. Aggregates only, never a re-hosted roster, and the non-partisan red line is doing real work here. The overlap finding (the trackers converged, and a two-source capture-recapture is invalid on nested lists) is the part a stranger can check.

Skipped 2026-09-18: The newsletter is not going out. Owner's decision, 2026-09-18, four days past the 2026-09-14 slot. This slot's primary channel was the first newsletter send, and there is no send to log, so it is recorded as a skip rather than backfilled with a witness for something that did not happen. The work itself is not affected: the research surface and its record-level overlap layer are live, and the derived lanes on this slot (indexnow, site-feed, correspondence) are not what this entry speaks for. First of the three skips the cap allows; two remain.

release / via correspondence / then indexnow, site-feed, newsletter

Reader three, and the slot most likely to convert into something concrete. Goes to village staff and the clerk directly, because that is where this audience is. The FOIA log is the second artifact in the same send.

Skipped 2026-09-25: Owner's decision, 2026-09-25, one day past the 2026-09-24 deadline. The slot's send was a note to the Village Clerk carrying the thread-one release with the FOIA log beside it, and the owner is not sending any FOIA requests yet. The Clerk is also the Village's FOIA officer, so a note tied to the log would read as a request. The clerk note itself (release only, no FOIA content) is drafted in docs/pingree-grove/thread-01-correspondence.md and deferred, not sent, so there is no witness to log. The release and the public FOIA log stay live at /research/pingree-grove, and the three FOIA drafts stay at status draft. Second of the three skips the cap allows; one remains.

release / via newsletter / then indexnow, site-feed

Thirty relationships, thirty disclosure sources, twenty-four of thirty deep-linked to a primary document and the other six honestly null. The published /audit ledger is what makes it citable rather than merely interesting.

Skipped 2026-10-02: Owner's decision, 2026-10-02, the day the gate marked the 2026-09-28 slot missed. The slot's primary channel was a newsletter send, and the list's only confirmed subscriber is the owner, so a send would have reached no reader and there is no witness to log. The work itself is not affected: the dependency graph and its /audit citation ledger are live at /research/one-load-bearing-company, and the derived lanes on this slot (indexnow, site-feed) are not what this entry speaks for. Third of the three skips the cap allows; none remain, so the later slots whose primary channel is the newsletter need a replanned channel before they come due.

release / via social / then indexnow, site-feed

Replanned on 2026-10-07, inside this slot's due window. It was planned as an arXiv submission, which cannot be made by its deadline: since 2026-01-21 arXiv requires an author with no prior papers in the category to get a personal endorsement from an established author there, and its staff cannot give one. So the slot sends what is finished instead: SAGEN-Bench v2, pre-registered on OSF, CC0, with a verifier that re-derives every verdict from the raw record. Goes to Mastodon, Bluesky, Nostr and the Telegram channel with one finding: two of the four registered hypotheses failed, and the release publishes them as failures. The arXiv submission waits for an endorser and belongs to the next band. The atom id keeps its original name because ids are stable.

Skipped 2026-10-09: Owner's decision, 2026-10-09, one day past the gate marking the 2026-10-05 slot missed. The slot was replanned on 2026-10-07 from an arXiv submission to a social post, because arXiv now needs a personal endorsement for a first paper in the category. The owner has chosen to hold the announcement and send it together with the arXiv submission once an endorser is found, rather than announce the release twice. No post went out, so there is no witness to log. The work itself is not affected: SAGEN-Bench v2 stays live at /research/sagen/bench, pre-registered on OSF, with its verifier. Fourth skip; the cap was raised from 3 to 4 to hold it (see SKIP_CAP), and the calendar needs replanning before the next one.

pointer / via correspondence / then indexnow, site-feed

Goes to TRAC (the Transactional Records Access Clearinghouse), the register's most-cited publisher with seven of its source rows, and so the reader best placed to say whether its figures are read correctly. No hashed release, so it travels as a pointer with one finding attached: the denominator switcher, where swapping a counting convention flips a verdict in place. A letter rather than a post on purpose: an enforcement claim seen out of context reads as a side, and symmetric scrutiny is the whole defense. If the October review has not landed, the letter states the 2026-07-24 run date.

Skipped 2026-10-16: Owner's decision, 2026-10-16, the day the gate marked the 2026-10-12 slot missed. The slot was a letter to TRAC pointing at the enforcement claims register with the denominator switcher as its one finding. The owner judged the letter too political to send right now. No letter was sent, so there is no witness to log. The work itself is not affected: the register stays live at /research/ice-claims with its 2026-07-24 run date. Fifth skip; the cap was raised from 4 to 5 to hold it (see SKIP_CAP). The replan the week 4 skip asked for is now overdue, and the next slot to come due with no entry is its deadline.

release / via social / then indexnow, site-feed

The most human thing in the catalog: the legibility argument turned on its own author, 902 games, 95.1 percent carrying the same unexamined third move. Goes to Mastodon, Bluesky, Nostr and the Telegram channel with that number, because it survives being seen out of context. Aggregates only: no opponent's username appears in any post.

pointer / via social / then indexnow, site-feed

Four threads of recovery legibility, and the empirical companion to the Wartime Edition essay. Goes to Mastodon, Bluesky, Nostr and the Telegram channel with one figure, the headline gap between compensation disbursed so far and need assessed (not what is owed), read off the page on the send day. If the quarterly refresh has not landed, the figure goes out with its 2026-06-22 date stated. A post rather than a letter because no single reader has been identified; the earlier plan to reuse the week 1 list did not fit, since that list is language-technology researchers.

pointer / via social / then indexnow, site-feed

OUT BEFORE FRIDAY. Departure is 2026-11-06 and this slot's Monday is 11-02, so it ships at the start of its week rather than the end. Goes to Mastodon, Bluesky, Nostr and the Telegram channel, because it is self-contained and a post owes no reply, which is what a slot before two dark weeks needs. One number: 117 of p53's 393 residues fall in the model's very-low confidence band, and 91 of them sit in the two disordered tails. The structure data is credited CC-BY 4.0 to EMBL-EBI and DeepMind.

dark

2026-11-09 to 11-15. Abroad. Nothing is scheduled, nothing is owed, and the missed-slot gate does not fire, because a dark week is a decision rather than a gap. This is the difference between a calendar you can keep and one you abandon in week ten.

dark

2026-11-16 to 11-22. Return lands 11-17 with a Taipei layover behind it. Publishing on the two days after a fifteen-hour flight is the kind of plan that gets a calendar thrown away, so the week is dark on purpose.

pointer / via social / then indexnow, site-feed

Re-entry slot. Deliberately the lowest-effort send in the band: a finished essay, a pointer, one post to Mastodon, Bluesky, Nostr and the Telegram channel, no new assets. Coming back from two dark weeks is the second most likely place to drop the habit. The one number is the inventory, about 49 unexercised options, and the post stays with the method (what an acquisition optimizer would cut, keep or monetize) rather than taking a side on any of it.

pointer / via correspondence / then indexnow, site-feed

Finished, published to the site in July, never sent anywhere. Goes to the author of the June 2026 Vectoral field report the essay is built on, because the analysis is a reading of that reporting and its author is best placed to say where it goes wrong. The page's red line holds in the letter too: no relay is named or linked.

pointer / via social / then indexnow, site-feed

Paired on purpose. The essay is the argument and the cross-paper Claims Ledger is the receipt. Goes out as a two-post thread on Mastodon, Bluesky, Nostr and the Telegram channel: one claim from the essay (Ukraine's property register is about 40 percent complete and mostly post-2013), then the Claims Ledger row that sources it. This is the send that demonstrates the house method rather than describing it.

pointer / via correspondence / then indexnow, site-feed

Published to the site in July 2026 and never distributed. Goes to the two authors of the AI Fluency summary document the statement names as its base content, the two readers who can say whether this close reading of it is fair.

digest / via social / then indexnow, site-feed

Closes the backlog band. Nothing new is written: a thread of at most eight posts on Mastodon, Bluesky, Nostr and the Telegram channel, one per public (indexed) hashed release, each carrying one number from its own page and a link home. The counts come from the catalog on the send day rather than from this note, because they have already moved since the calendar was drafted. The last slot that can be filled without making anything.

episode / via correspondence / then podcast-feed, site-feed, social / needs backlog-band-complete

The standing advice for a year has been to ship one episode before building more infrastructure. Since that advice, two complete show architectures were built instead. This is the slot that stops being ironic. The guest page is the deliverable; the audio is exhaust. Guest-driven: if no guest has said yes by 2026-12-14, the slot is replanned in the open while it is still scheduled, because the skip cap is spent.

Displaces: Two backlog slots' worth of effort, and concretely the /store and /reading distribution that would otherwise have filled weeks 17 and 18. Deferred to here rather than earlier because it is the only item in the picture that is new production work, and the backlog band exists to stop new work from displacing finished work.

submission / via directories / then podcast-feed / needs between-systems-ep1

Displaces nothing: it is unlocked by week 17 rather than competing with it. All seven directory edges in the graph are correctly marked planned today because a directory cannot ingest a feed whose only item has an empty enclosure. One episode makes seven edges legal at once.

episode / via social / then legibility-feed, site-feed / needs between-systems-ep1

The show opens the only honest way a critique show can, by scoring its own house first. Recording it is new work; everything it says was written months ago. Once the episode is in the feed it goes to Mastodon, Bluesky, Nostr and the Telegram channel with one of the three scores it gives the house. A self-audit has no outside subject, so there is no one to write to first.

Displaces: A new research entry. The script, rubric, steelman and both receipts already exist and have since July 2026; only the audio is missing, so this is closer to backlog than the kind marks it.

audio / via social / then radio, site-feed

Closes the gap the inventory found: narrated audio exists on catalog entries 01 through 09 and almost none of the 2026 run, which is the strongest work, has anything for the radio surface or any audio-shaped channel to carry. Goes to Mastodon, Bluesky, Nostr and the Telegram channel with a link to the radio and the count of entries that gained a voice. Unratified already has its own narration and is not in this batch.

Displaces: New writing. Nothing is authored here; prose that has existed since 2026 finally gets a voice.

release / via correspondence / then indexnow, site-feed, social

Thread two, the property-tax split across the nine taxing bodies on a Pingree Grove bill, shipped as a hashed release and has not been sent. Reader three, public records only, and the thread most likely to matter to the people who live here. The letter goes to a reader named by role who is not the Village's FOIA officer, because the Clerk is also the Village's FOIA officer and a note to that office can read as a records request (the week 3 skip); if no such reader is identified by 2027-01-11, the slot moves to social while it is still scheduled.

Displaces: Nothing new. Thread two is already published, so this slot sends finished work.

digest / via social / then indexnow, site-feed

Twenty-one reusable systems in five stacks with live usage metrics, the self-checking evals, and the architecture map. Nothing new is written: a three-post thread on Mastodon, Bluesky, Nostr and the Telegram channel, one surface and one number per post, the first naming the 21 systems in five stacks. This is the one slot aimed at reader four, the only reader who needs reach. Every post is about this site's own systems and nothing else.

release / via correspondence / then indexnow, site-feed, social / needs ukrainian-frequency-v1

Twenty-two weeks after version one, with whatever the correspondence from week 1 taught. A release that answers its own readers is the strongest evidence the correspondence thesis is right, and the strongest evidence against it if nobody replied.

Displaces: A fourth podcast episode. Reader one outranks the show, and a second version of the list is worth more to them than another hour of audio is to anyone.

episode / via correspondence / then podcast-feed, site-feed, social / needs between-systems-ep1

The referral chain in the catalog exists for this: a guest who suggests the next guest is the mechanism, and it cannot start until episode one exists.

Displaces: One backlog-style digest. Guest-driven, so the date is a commitment to ask, not a commitment to publish. The skip cap was spent by week 4, so if no guest has said yes by 2027-02-01 this slot is replanned in the open while it is still scheduled, never skipped and never quietly re-purposed.

release / via social / then indexnow, site-feed

Catalog entry 13 has been In Progress since June 2026. Finishing something already begun is the cheapest new work available. Goes to Mastodon, Bluesky, Nostr and the Telegram channel with one number, the five deployed systems it reads: three identity layers (Aadhaar, Estonia's eID and X-Road, MOSIP) and two payment rails (UPI, Pix), because the explorable holds up out of context. The Aadhaar case is re-verified the week it ships.

Displaces: Nothing. It completes a catalog entry that has been In Progress rather than starting a new one.

review

No atom leaves. The ledger is the evidence: how many slots published inside their grace window, how many were logged skipped and why, whether correspondence produced replies, and whether the reader ranking survived contact. The calendar expires after this week, and the gate fails when it does, which is the mechanism that forces this review to actually happen rather than the schedule quietly running out.