The year's totals

The ledger carries twenty-two platform objects and eleven directory pages, and those two numbers describe different kinds of coverage rather than two views of one list. An object exists in this ledger because the site decided to track that name. A directory page exists because a public catalogue lists the platform. A name can have the first and not the second, which is exactly what makes the second worth counting separately. Reading the two as one figure hides the gap that matters to anybody moving a file.

The address counts for the single period this site measured are narrower still. Ninety-three addresses were requested in the 2026-W39 window: thirty-one wrapped rows taken from public sheet pages and sixty-two rows drawn from a directory listing. All ninety-three came back in the same state, with no row recorded as gone, moved, or shielded. A first-period distribution that flat is a description of one moment, and it gets repeated here as a description rather than as a result about how links behave.

Two more figures belong in the totals because they explain what the first two cannot. One hundred and thirteen addresses were dropped when the host stopped accepting connections, and they were excluded rather than recorded as unknown, so the published sample is smaller than the number of addresses the work touched. The checking setup was deliberately narrow as well: eight seconds per address, no account credentials, and no more than two requests running against one host at a time. Both figures shape what the yearly tally can support.

Sorting the twenty-two objects by recorded status gives four groups. Two entries are closed under one operator: ACBuy and AllChinaBuy. One is closed and kept as a historical reference from 2024: PandaBuy. One sits as a legacy record: Mulebuy. Nine are recorded as accepting orders, and the remaining ones carry an unverified status, meaning this ledger has no reading for them rather than a poor one. A year in four groups is not much of a story, and it is the honest shape of the evidence.

Against the year before

The comparison cannot be made from this site's own numbers, and that is worth saying before anything else in this chapter. There is no earlier period on file, no prior ledger snapshot, and no measurement from a previous year to put beside 2026-W39. Anything phrased as a year-on-year change here would be a comparison between one measurement and a set of public announcements, which is not the same kind of comparison. The gap is the finding, not an inconvenience on the way to one.

What the public timeline does support is a short list of dated events, and each of them is a statement rather than a measurement. PandaBuy closed permanently in 2024 and has stayed closed, which makes it the year-before reference point available to anyone looking at this category. In 2026 this platform posted a security and compliance notice on 2026-06-01 that mentioned disrupted addresses, slower pages and some cancelled orders, then stopped purchasing and redemption from 2026-06-12 through the rest of June while warehouse and parcel work continued under the same notice.

The second half of that timeline is quieter and more useful. A staged return announced for 2026-07-01 did not restore ordering. Around 2026-07-10 a community representative said ordering had not returned and gave no new date. By August, third-party records showed the storefront offline, while the platform itself never announced a permanent closure. Notice, suspension, a missed restart date, silence, and an empty storefront: those are the facts of the year, and none of them is a number this site produced.

The difference in method is the part most comparisons skip. This site counts addresses that answered on a stated day, one row at a time, with the failures kept in the record. Public records count something else: operator statements, closed storefronts, and whatever a catalogue still lists. A shutdown announcement and a column of unreachable addresses can describe the same event through different instruments, and neither one converts into the other. Both belong on the page, labelled for what they are.

What changed structurally

The first structural change is that a closure can now happen without ever being announced. PandaBuy closed in 2024 with a public end to it. The 2026 case produced a compliance notice, a suspension, a restart that did not arrive, a statement that ordering had not returned, and then a storefront that stopped serving, with no announcement of permanence at any point. A reader waiting for a final notice waited through the whole sequence and is still waiting. The record that exists is a trail of partial statements rather than an ending.

The second change is in who can say whether a platform is operating. With no closing notice, the directory page becomes the load-bearing public signal, and this ledger tracks that coverage as a field for exactly that reason. Eleven of the twenty-two objects appear in the catalogue; the rest are names this site follows without a page to read. A category where the answer to is it still open comes from third-party listings is a category where the useful question shifts from announcements to coverage.

The third change is what turned out to be durable. Wrapper addresses stopped answering, and the identifiers inside them did not, which moves the value of an old sheet from the links to the digits and the dates. That shift shows up in the fields this ledger records, where address handling and check dates carry more weight than anything descriptive. It also shows up in what people keep: the file with dated rows survives a platform change, and the file of pretty links does not.

The fourth change is the cost of being certain. Certainty has not disappeared from this category, it has moved to whoever holds the primary record, and for most readers that means an account page or a dated public notice rather than a summary anywhere else. A site like this one can publish what it measured, on the day it measured it, and list what it deliberately left out. That is a smaller product than a verdict, and it is the only one the evidence supports.

A forecast for next year

No forecast follows, and the absence is deliberate rather than coy. One period of ninety-three addresses supports no statement about where this category goes over twelve months, and a confident direction would be a guess wearing the clothes of a total. What the method supports instead is a short list of things to watch and the order to check them in, so a reader can reach a private reading without depending on a summary written by anyone else.

Watch the directory coverage first, because it is the cheapest public signal in the set. Count how many platform objects have a page when the next window is read, and compare that count rather than the names. A count that falls is the clearest available sign that the category is consolidating, and a count that holds while names change underneath says something different about the same year. Keep the day that count was taken beside it, and next year the comparison exists.

Watch the storefront of the closed platform second, and watch it on a dated schedule rather than on impulse. The specific things worth a note are a return of purchasing for accounts generally, a statement of permanent closure, or a change in directory coverage, in that order of usefulness. Anything short of those, including one responsive request or a rumour about a partial return, is not a reading worth writing down. Numbered and dated notes beat a memory of the week something happened.

Watch the address distribution third, since it is the only figure this site produces from its own work. Whether the next window keeps ninety-three rows in one state, and how many rows land as blocked or missing, is a fact worth having beside this period's flat reading. Keep the dates attached when the comparison is made. The useful posture for the coming year is not prediction but a repeated check written down on the same day each time.