Three products arrived, and one badge had been wrong for three days

Three products joined the catalogue this week: Vela, a language teacher that keeps every language you learn on its own separate record; Auren, one place for cycle, mood, pain, sleep and food; and Forgotten Town, a merge game about repairing a town nobody came back to. Each of them arrived with real screens, a real identity and a demo you can run on this site.

The interesting part of the week was not any of those. It was the re-audit of the products that were already here.

The badge that had been wrong for three days

This site said Pivumo was COMING SOON, and that its audio was finished but the app had not been submitted. That had been true when it was written. It stopped being true on 22 September, when Pivumo went live on the App Store — and nothing about this repository changed at that moment, so nothing here noticed.

It is the second time the same mistake has been made in the same direction. An audit in September concluded that no product was publicly available, while three had already been listed for four days. Both times the method was the problem rather than the diligence: the question "is this product public?" was being answered by searching the code, and the code is not where the answer lives.

So it is a script now. It asks Apple for every bundle identifier, lists every app on the developer account, and requests each package page from Google Play. It runs before any release claim on this site is edited, and it ran again immediately before this was published.

What a re-audit is actually for

A catalogue page is not a description of a portfolio. It is a set of claims about ten separate pieces of software, each of which is being changed by somebody, on a schedule that has nothing to do with this website. Every one of those claims decays. The useful question is not whether the page is accurate today; it is which kinds of inaccuracy the page can develop without anybody noticing.

• A status badge that has fallen behind the store. Now impossible to hold: a LIVE badge and a public store listing are asserted to be the same fact, and the build fails if they disagree.
• A count in a sentence. "Seven products" read perfectly well on a page listing ten, which is exactly why nobody counts. Every number this site states about its own catalogue is now computed from the catalogue, and a test fails the build if one is typed by hand.
• A product added without a demo. The strongest sentence here is that everything can be tried. Adding a product is an afternoon; building its demo is not. That claim is now derived rather than written down, and it failed three times this week before it passed.
• A number quoted from a plan instead of a build. Forgotten Town has a hundred orders and five areas, and both figures are counted out of the content file the shipped game loads, not out of the document that proposed them.

Two products that will not say what they were said to say

The brief that introduced Vela to this site described production and security work already in place, an on-device teacher model disabled after a quality evaluation, and a server-backed guided mode as the current product experience. None of that is in the Vela project. Its own handoff document states, in terms, that there is no backend, no account, no persistence, no payment, no AI and no speech recognition, and a search of its source for any HTTP client, model provider or API key finds nothing at all.

Vela is a genuinely finished interface with ten languages, a real architectural idea and thirty-three passing tests. That is a good thing to be, and it is what its page says. The rest is not published, because it could not be verified — and a studio that prints the flattering half of an unverified brief has no business asking anyone to believe the rest of its site.

Auren went the same way for a smaller reason. It was described with photo-assisted food analysis and wellness recommendations. Both exist — as interface flows, explicitly simulated in this build. So its page says flow, not analysis, and its demo quotes the app’s own carefully hedged wording rather than paraphrasing it. That hedging is the product: an app in this category that says "a personal forecast needs reliable patterns first" is making a harder and more useful promise than one that just forecasts.

The demos got a better ladder

With ten demos the old labels stopped working. "Playable" was carrying both the demos that run the product’s own source and the one that plays its shipped audio, and those are not the same claim. There are four rungs now — the product’s own code, the product’s own material, a recreation, and a walkthrough that computes nothing — and each demo says which one it is standing on.

Forgotten Town’s demo is the clearest case for the middle rung. The merge loop is written for the browser, because a Flutter and Flame build cannot go in a web page at a weight this site would accept. Everything it operates on is the game’s: the board, the chain, the sprites, and order h1 with the item it asks for and the twenty coins it pays, read out of the file the shipped build loads. It is not the game. It is the game’s first thirty seconds, made of the game’s own parts, and saying exactly that is worth more than calling it something it is not.