vela

Proof · worked example

A worked Correction

One event, end-to-end. The substrate seen working.

Three real papers (Halliday 2016, Nation 2019, Montagne 2020). Four signed events. One Correction propagating through a dependency graph to three downstream objects. Reconstructed from the actual literature; the substrate-mediated propagation shows what would have happened if Vela had existed in 2020. Confidence numbers are illustrative; the event shapes and propagation paths are the load-bearing claim.

2026-01-08 · Finding

The original finding lands.

Halliday and colleagues post-mortem work on hippocampal pericyte degeneration is compiled into a Vela finding by an agent and accepted by a reviewer.

vf_4a91ce… · supported · conf 0.62

Pericyte loss precedes cognitive decline in early Alzheimer's disease.

scope
human · sporadic AD · prodromal stage
evidence
1 EvidenceObject · postmortem hippocampal histology
cite
Halliday et al., J Cereb Blood Flow Metab 2016
signed by
reviewer:demo-curator-1 → vrev_71f2… (illustrative; no real PI signed this)

At this point the constellation has one node here. No dependents yet.

2026-04-14 · EvidenceObject · ConfidenceUpdate

New evidence attaches. Confidence rises.

Nation and colleagues' longitudinal MRI study showing blood-brain-barrier leakage as an early biomarker of cognitive dysfunction is added as a second EvidenceObject. The reviewer signs a ConfidenceUpdate; the same finding now carries 0.78.

before 0.62
after 0.78
ConfidenceUpdate vrev_3c7e91…
target
vf_4a91ce…
delta
+0.16 (0.62 → 0.78)
cite
Nation et al., Nature Medicine 2019
signed by
reviewer:demo-curator-2 → Ed25519 over canonical preimage (illustrative)
signed at
2026-04-14T11:42:08Z

Same node, brighter. Two EvidenceObjects attached.

2026-06-02 · Correction

A 2020 paper narrows the scope.

Montagne and colleagues' 2020 paper showing the BBB breakdown is concentrated in APOE4 carriers lands. A reviewer drafts a Correction event and signs it. The scope narrows from sporadic AD to APOE4-positive sporadic AD; confidence on the broader claim drops; a new child Finding spawns for the APOE4-positive subgroup.

Correction vrev_b0144a…
target
vf_4a91ce…
scope before
human · sporadic AD · prodromal stage
scope after
human · sporadic AD · prodromal stage · APOE4-positive
confidence on broader claim
0.78 → 0.41
spawns
vf_d92710… (APOE4-positive subgroup, conf 0.74)
cite
Montagne et al., Nature 2020
signed by
reviewer:demo-curator-3 → Ed25519 (illustrative)
signed at
2026-06-02T16:19:33Z
broader 0.41
APOE4+ 0.74

The constellation now has two nodes here, linked by a parent-of relationship. The original is dimmer; the child is brighter; the seam is visible.

2026-06-04 → 2026-07-11 · Dependency notification

Three downstream objects find out.

The dependency graph notifies three objects that previously rested on the broader claim. None of these teams talked. None of them needed to. The substrate carried the correction.

  • + 2 days · biotech · Cambridge MA

    A target hypothesis at one neurovascular biotech updates its priors: the program redirects from a sporadic-AD pan-cohort design to an APOE4-stratified Phase Ib. The portfolio review the next Friday takes thirty minutes instead of three weeks.

  • + 39 days · review article · Osaka

    A review article's headline claim picks up a margin gloss: the broader pericyte-loss-precedes claim is annotated with the APOE4 stratification before the proof galley is signed. The reader sees both lines on first read.

  • + next protocol-amendment cycle · trial · Toronto

    A Phase II inclusion criterion that depended on the broader claim is flagged for review at the sponsor's next protocol-amendment window. The flag carries a back-link to vrev_b0144a… so the data-monitoring committee, the IRB, and any future auditor see the same reasoning when the amendment is drafted. Trials run on their own clocks; the substrate does not pretend otherwise.

The biotech moves in days, the review article on the next galley cycle, the trial on its own protocol-amendment window. Each downstream object updates on the clock its institution actually runs on — but every one of them starts from the same signed Correction event instead of rumor or a chance encounter at a conference. In the world the essay describes, the same loop typically takes five to twelve years, when it closes at all.

This is one Correction event. The substrate carries millions.

Every Finding above is a content-addressed object. Every event is signed. Two implementations replay this same sequence and reduce to the same hash, or one of them is wrong. That is the only invariant the protocol promises; it is enough to make the field navigable.

Open the workbench → · Read the spec →