When an app listing says “supports foldables,” what does that actually prove?
It might mean the app fills a larger window. It might mean one developer tested one device. It might describe a two-pane layout, a cover-screen feature, or only a release-note keyword. None of those automatically tells us what happens when the device is folded, unfolded or half-open.
While building FoldKit, I ended up treating foldable compatibility as an evidence problem rather than a yes-or-no field.
Three states, not one badge
A useful source may cover one or more states:
- Folded: cover-screen or compact-window behavior
- Unfolded: expanded layouts, multiple panes and larger-window behavior
- Half-open: tabletop, book, hover or hinge-aware arrangements
An app can behave well in one state and still have unknown behavior in the others.
Four different kinds of information
I also needed to keep these separate:
- Listing metadata — identity, version and storefront information
- Developer claims — features described by the maker
- User reports — individual experiences with device and version context
- Scoped posture evidence — a source that explicitly describes a result for a particular state
Collapsing those into one score creates false precision. If no source explicitly covers a posture, FoldKit leaves it untested.
Context changes the meaning
A user report without a device, OS version or app version may still be useful, but it cannot establish a platform-wide defect. A reply that reports a fix matters. Conflicting results matter. A system setting that resolves the issue changes how the report should be summarized.
The goal is not to discard imperfect evidence. It is to keep its scope attached.
Human review around agent collection
Agents help discover and summarize candidate sources, but every new Kit entry starts as a draft. The intake pipeline checks structure, duplicates, source availability and repository metadata. Production excludes drafts until a human review approves them.
The same boundary applies to the product itself: FoldKit does not install or run the apps and tools it indexes, and it does not present author claims as FoldKit testing.
The first public release contains 45 guides and tools, 30 app profiles and 19 source-linked user reports. You can explore the evidence model at foldkit.me.
I would be interested in how other developers model compatibility claims. What would you require before changing a posture from “untested” to a reported result?
Top comments (0)