John Malone · Human-directed, AI-drafted

Brief mode: condensed view. Switch to Full for persona notes and full analysis.

Context: A useful product demonstration shows what the system can answer, what it used, and where the available evidence runs out. Rowan, Seton, Campion, and Devil’s Advocate are AI personas used as editorial lenses; John is the human author.

Leave the Missing Data in the Demo

A useful product demonstration shows what the system can answer, what it used, and where the available evidence runs out.

A demo gets easier the moment you allow yourself to invent the records.

The question can be perfectly phrased. The answer can be sitting in exactly the right document. Every search can return something flattering. Even the loading spinner can develop excellent manners.

The problem arrives when your actual life refuses to resemble the fixture.

While working on an Alcove Home demonstration, I kept coming back to a simple constraint: use real material, and show what the available material can support. I want to see useful answers from my own digital life. I also want to see where the collection runs out.

Show the edges of the collection

That makes a less obedient film. It makes a more useful engineering exercise.

A personal search system can fail for several different reasons. The record might never have been imported. It might be cataloged without its contents being searchable. The searchable copy might exist while the original is unavailable. Or the evidence might be present and the retrieval simply miss it.

Those cases deserve different explanations. “No results” is often too small a container for everything it is being asked to conceal.

The distinction matters before the first query. If I have selected a small collection, the interface should tell me what is in it. A search box sitting above a partial collection should not imply access to every account, photograph, and document I have ever accumulated. An inventory of file locations is useful, but it is not the same as searchable file contents.

I want that coverage to be visible in the demonstration, even when it makes the product look unfinished. It tells us what the next piece of work actually is.

Show the assistance, too

Real material also produces more interesting tests. In one prototype session, an ordinary request to find the photograph behind a set of AI artworks turned into a search using remembered context, local metadata, and visual comparison. A likely match was found, alongside a near-duplicate. The remaining uncertainty concerned which file had actually been used to generate the artwork.

That was an agent-assisted session, not a demonstration of an automatic Alcove Home photo-search feature. Calling it one would turn an informative prototype into a misleading product claim. Showing what worked and what still required assistance gives us a much better basis for deciding what to build.

Test the experience you are claiming

The same discipline applies to offline access. A page that remains on screen after the connection drops can look convincing in a video. The test I care about is more demanding: reload, ask a new question of the prepared local collection, and open the supporting evidence.

That is the experience we have set out to verify. A device can only search and display what it actually has available. If an attachment was never downloaded, the interface should say that rather than leave the viewer to infer that a cached title represents a cached document.

There is room for good editing. Nobody needs to watch every wait, every scroll, or every mistaken click. Editing becomes a problem when it changes the claim: removing the missing-data state, disguising manual assistance, or presenting a proposed interaction as something the product already does.

I would rather show a short, real success and a clear boundary. Here is the question. Here is the answer the selected records support. Here is the passage or image behind it. Here is what we cannot establish yet.

An honest gap can be productive. It might tell us to improve retrieval, prepare a local copy, or bring in another authorized source. It might also tell us that the answer is unknown. Better lighting will not resolve that last one.

The point of the demonstration is to find out whether this is becoming useful. When the evidence runs out, leaving that moment in the frame is part of the result.

Conversation notes: September 26, 2026. Published September 28, 2026.