feat(lifeboard): provenance — every fact can say where it came from - #317
Open
ivanmkc wants to merge 1 commit into
Open
feat(lifeboard): provenance — every fact can say where it came from#317ivanmkc wants to merge 1 commit into
ivanmkc wants to merge 1 commit into
Conversation
Answered from the log rather than from a side table. The log already records, for every change, what it was derived from, who made it and what it replaced -- recorded at the moment the fact was created, because nothing else was possible. That is what makes provenance work at all; a derivation bolted on afterwards is a guess, and a guess about where a vaccination date came from is worse than nothing. Derivation records add only the two things the log cannot infer: which kind of step it was, and how confident a model was. A source that has changed is flagged, never silently re-read. Re-running an extraction against a changed document behind someone's back is how a date quietly becomes a different date. Lineage is drawn as a flow board because the thing people need is the shape -- that a date came from one email through one extraction, or from three documents through a merge and a correction. A list of steps reads as bureaucracy. Deep links live in a table as data. Every connector addresses a document differently and the formats change without warning; a broken link should be a one-line fix. Drive gets the page number, Gmail opens the thread. Boards reach the rest of the app through an href prefix the shell intercepts. A board is data and cannot carry a click handler -- giving it one would mean the agent authors behaviour rather than a picture. Anything else in an href is an ordinary link and leaves the app, which is what a source link should do. Pairs a derivation to the entry it explains by sequence number, not timestamp: a test caught two unrelated writes landing in the same millisecond, which mislabelled the history of the fact someone was trying to check. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This was referenced Sep 2, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The stack
localStoragefor the test suite on Node 25@ivanmkc/termchart-canvas; injectable interact transportEvery branch in the stack was checked out on its own and verified independently:
npm install,tsc --noEmitfor every package,npm testacross the workspace, and the offline e2e where it exists. All nineteen are green — no branch depends on a later one to build or pass. Full viewer e2e (11 suites, 106 assertions) was run on #312 as the behaviour-preservation evidence for the extraction.Stack position: 7 of 10. Base:
lifeboard/05-packs.Every displayed fact can answer "where did this come from?" and take you to the document.
Answered from the log rather than from a side table. The log already records, for every change, what it was derived from, who made it and what it replaced — recorded at the moment the fact was created, because nothing else was possible. That is what makes provenance work at all; a derivation bolted on afterwards is a guess, and a guess about where a vaccination date came from is worse than nothing. Derivation records add only the two things the log cannot infer: which kind of step it was, and how confident a model was.
A source that has changed is flagged, never silently re-read. Re-running an extraction against a changed document behind someone's back is how a date quietly becomes a different date.
Lineage is drawn as a
flowboard because the thing people need is the shape — that a date came from one email through one extraction, or from three documents through a merge and a correction. A list of steps reads as bureaucracy.Deep links live in a table as data. Every connector addresses a document differently and the formats change without warning; a broken link should be a one-line fix. Drive gets the page number, Gmail opens the thread.
Boards reach the rest of the app through an href prefix the shell intercepts. A board is data and cannot carry a click handler — giving it one would mean the agent authors behaviour rather than a picture. Anything else in an href is an ordinary link and leaves the app, which is what a source link should do.
A bug worth noting: derivations were paired to entries by timestamp, and a test caught two unrelated writes landing in the same millisecond — which mislabelled the history of the fact someone was trying to check. They are paired by sequence number now.
Verified: 257 unit tests, including the gate's example — a vaccination date traced back to a PDF page and an email paragraph, corrected, and flagged when the source changes.
🤖 Generated with Claude Code