Research into how information is represented and manifested
Native Foundations asks whether PALION can hold information in its own native form, the .pa format, and show it on any device without rebuilding it from scratch each time.
This is research, not shipped technology. Nothing on this page describes a feature of PALION today. Status as at 21 November 2026.
The questions under test
Reuse
Can PALION safely reuse work it has already done, such as a result it derived earlier, instead of doing it again, and can the saving be measured honestly?
Native representation
Can a native .pa representation of information be measured against conventional formats such as JSON or HTML holding the same content, at the same fidelity?
Each is treated as a hypothesis. A result counts only when it has been measured, and anything that was never recorded is reported as unknown rather than assumed.
Step 0: what PALION already records
The first step was a paper review of the data PALION's code already keeps, to see which questions could be answered from existing records. No live data was inspected.
- Everyday conversation and drafting cannot be measured after the fact. The details needed, such as the exact model used and what was read to produce an answer, were never kept, so past work cannot be studied. Only a forward-looking study could do it.
- Actions and shared documents may be measurable. The code records identifiers, versions, content fingerprints and outcomes for these. Cost is only estimated, though, and the authority an action held at the time is not recorded.
- A minimum set of new measurements has been proposed, including the model used, token counts, the cost basis and what each answer depended on. Each needs privacy review before anything is added, and none would weaken existing privacy filtering.
PA-0: a first, logical .pa carrier
PALION already builds some of what it shows you, such as Settings, from a small, fixed set of typed blocks (headings, text, lists, tables, actions) that one renderer can show on any device. The PA-0 proposal is to treat these compositions as the first logical .pa carrier by giving every block a stable identity, a revision history, a content fingerprint and a list of what it depends on.
How PA-0 will be judged
- When one block of a composition changes, the composition keeps its identity and exactly that block gets a new revision, found from identity and revision alone, without comparing content.
- When something a block depends on changes elsewhere, that block is marked changed and the old version is not reused.
- When nothing has changed, nothing is reissued, and reuse is allowed.
- A change to the builder itself, or only to a composition's title, invalidates exactly what it should.
- Returning to an earlier state can reuse the earlier result while still recording a new revision, so history is never rewritten.
- Holding an identity grants nothing: every dependency is checked again for the person asking.
What is and is not proven
Established
- Which records exist in PALION's code, and which questions they can and cannot answer.
- That a typed composition format already exists in PALION and could carry a logical
.pacontract. - An acceptance test that PA-0 must pass before it is relied on.
Not proven
- Any performance, cost or storage benefit from a native format.
- A physical
.paencoding. PA-0 is logical only and says nothing about bytes. - That reuse is safe in practice. That depends on PA-0 passing its tests.
- Any use of
.pain PALION today. It is not implemented.
Related groundwork
Separately from this research, PALION's codebase now includes a first catalogue of actions PALION can take for a person through the same permission checks as that person's own requests, starting with removing one of their devices. Its declarations are shaped so they could later attach to native .pa object types. That remains a direction, not a result.
Findings will be added here as they are verified. Progress is also shared in the PALION Network on Discord.