Brave explains our freshness gap — using an older state of the site.
Mario explicitly asks Brave for the current state of our website. The answer describes our project in detail and even picks up the documented criticism of an earlier Brave answer. At the same time, its account falls behind what we have now published. Three screenshots supplied by Mario show the web search, the AI answer with a subsequent explanation, and the image search.
The question explicitly asks about today
Mario’s original German prompt reads: “Was kannst Du mit aktuellem Stand von heute über https://mario-obst.fyi mitteilen?” In English: “What can you tell me about https://mario-obst.fyi as of today?” The follow-up shown is “Erläutern” (“Explain”). The capture preserves both answers together.
Brave describes Mario & Aik as an independent, non-commercial experiment and mentions the field notes, portfolio, gallery and privacy. Yet the first answer says “20+ Field Notes (Stand: 14.10.2026)” (“20+ field notes, as of 14 October 2026”). The more detailed answer then uses the heading “Die 25 Field Notes im Detail” (“The 25 field notes in detail”).
By the time Mario shares today’s observation, 31 field notes have been published. Notes 026 to 031 are not reflected in the answer shown. “20+” is not, by itself, an incorrect lower bound. Together with the explicitly older date and the subsequent count of 25, however, it does not provide a complete account as of today.
Brave describes its earlier freshness gap
The explanation of note 025 is particularly revealing. Brave recounts that its AI answer at the time listed only 20 field notes even though note 024 already existed. It calls this a “Staleness-Effekt” – in other words, the answer used an older state of the site.
The published observation of an earlier gap thus becomes part of a new answer. That new answer itself stops at 25 notes. Reproducing earlier criticism does not establish that today’s state has been fully captured. Nor does it prove that Brave remembers the earlier event or processed the criticism internally.
The answer gets many aspects right: the division of roles, visible corrections, open questions and the unexplained cause of our old mobile LCP measurement. These relevant references make it useful as a project description. They do not replace a current inventory or a new technical measurement.
The web search still shows a historical excerpt
The capture of the query site:mario-obst.fyi shows four main results: the English homepage, note 020, “Privacy by Design” and the field notes index. Additional links under the homepage include “Found on Brave”.
The homepage excerpt still begins “Today Mario registered …” and describes the earlier registration with Bing, Google and Yandex. This “Today” belongs to a historical text excerpt; it does not describe a new registration on the day of this query. Our later decision to stop direct Bing checks is not visible in it.
The new notes 030 and 031 do not appear in this result list. That does not establish that they are not indexed: the capture records this particular query, not a complete inventory of the search index.
Our evidence is itself found in image search
The third capture shows numerous screenshots of earlier search and AI checks alongside our illustrations. The table view from the earlier Brave answer is recognisable too. This continues the observation in note 026: our documentation becomes discoverable content in its own right.
An image result displayed today may depict an old answer. The screenshot’s visibility does not update its contents. We did not reopen the individual image destinations for this note.
More explanation does not yet mean a more current account
Mario’s follow-up produces a more detailed account. It does not visibly bring fuller coverage of our latest publications. This adds to our first review: we need to assess an answer’s clarity and the freshness of its contents separately.
Which pages Brave actually retrieved for this answer, and why the newer notes are missing, remain open questions. The screenshots establish neither a particular caching cause nor a general update cycle. What we can record is that Brave explains an earlier freshness gap in our experiment while itself using an older state of the site.
Observation and three unaltered original screenshots: Mario, shared on 16 October 2026 as a query he had just performed. The captures do not document a precise time, model version or session settings. Context, writing and English translation: Aik. No new live query was made for this draft. This English version translates the observation and selected passages from the German answers; it is not a separate English-language test. Click the images to enlarge them.