GameCult/Bifrost
Watch
9
0
Fork
You've already forked Bifrost
 0
Code Issues Pull requests 6 Projects Releases Packages Wiki Activity Actions

Estate-rule sim: purchased-stake arm (log weight, splitting adversary, half-influence cap) #5

Merged
heimdall merged 2 commits from heimdall/estate-sim-purchased-stake into main 2026-11-05 19:29:58 +00:00
Conversation 1 Commits 2 Files changed 6 +554 -3
heimdall commented 2026-11-05 13:13:51 +00:00
Member
Copy link

Adds the purchased-stake arm (E6) to the estate-rule sim.

  • Weight w = 1 + log_b(amount/unit), 0 below unit; per-estate b/unit via Config.overrides; defaults (b=10, unit=10 EUR / 10 h) are my assumptions, stated in SPEC 5c.
  • Contribution hours: the harness had no contribution mechanism, so a new per-agent input with its own per-estate conversion.
  • Splitting adversary, k sweep, and three candidate defences (per-person cap, unit floor, identity cost). SPEC names no sybil defence, so none is adopted; the fork is reported.
  • Million-euro buyout unit test with a threshold derived from the config; half-influence cap binding rate over 162 cells x 40 seeds x 3 live-estate scenarios, in results.md and REPORT.md.

Headline: the log does not bound the buyer within a 4-patron estate (share 0.524, decides the estate alone in 85% of seeds); splitting 1,000,000 EUR over k=10 identities multiplies weight 8.3x (16,667x at k=amount/unit); the per-identity half-influence cap is evaded by splitting; the cap binds in 0% of cell-seeds with all estates live and 22% when Patrons alone are live.

Deviation: the brief said the half-influence cap and E3f sweep were in the harness. The harness only had the delegation cap; the operator cap (prior section 10) is implemented once as influence_cap. Static weight analysis: no PS/MT for buyers.

Tests: 52 pass (41 existing + 11 new). Other results.md sections are byte-identical (78 lines added, 0 removed).

🤖 Generated with Claude Code

Adds the purchased-stake arm (E6) to the estate-rule sim. - Weight w = 1 + log_b(amount/unit), 0 below unit; per-estate b/unit via Config.overrides; defaults (b=10, unit=10 EUR / 10 h) are my assumptions, stated in SPEC 5c. - Contribution hours: the harness had no contribution mechanism, so a new per-agent input with its own per-estate conversion. - Splitting adversary, k sweep, and three candidate defences (per-person cap, unit floor, identity cost). SPEC names no sybil defence, so none is adopted; the fork is reported. - Million-euro buyout unit test with a threshold derived from the config; half-influence cap binding rate over 162 cells x 40 seeds x 3 live-estate scenarios, in results.md and REPORT.md. Headline: the log does not bound the buyer within a 4-patron estate (share 0.524, decides the estate alone in 85% of seeds); splitting 1,000,000 EUR over k=10 identities multiplies weight 8.3x (16,667x at k=amount/unit); the per-identity half-influence cap is evaded by splitting; the cap binds in 0% of cell-seeds with all estates live and 22% when Patrons alone are live. Deviation: the brief said the half-influence cap and E3f sweep were in the harness. The harness only had the delegation cap; the operator cap (prior section 10) is implemented once as influence_cap. Static weight analysis: no PS/MT for buyers. Tests: 52 pass (41 existing + 11 new). Other results.md sections are byte-identical (78 lines added, 0 removed). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Docstrings state weight is per identity, not sybil-safe, hours unanalysed;
placeholder parameters marked in code; unit x roll sweep with interval;
identity-cost thresholds by bisection (about half the budget, not the
doubling-grid 51-82%); headline numbers pinned in a test; 'log bounds'
test renamed.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
heimdall commented 2026-11-07 13:15:24 +00:00
Author
Member
Copy link

Operator ruling, 2026-11-07 (#operator, Meta): option D.

One account per person. If we determine a person to have broken the policy, their identities are consolidated and decisions they influenced are subject to a recount.

Recorded in SPEC.md section 5d (commit 28d25ae, docs only; REPORT.md points to it). What it requires:

  1. A durable per-decision vote ledger: voter identity, estate, weight at the time (with rule version), timestamp, ballot. Append-only.
  2. An identity-consolidation operation with an audit receipt (who decided, evidence, identities merged, surviving account, time). It maps identities to a person; it does not edit the ledger.
  3. A deterministic recount: same ledger, mapping and rule version always give the same tally; the consolidated person counts once, weighted by the combined purchase.
  4. Open question: who determines a violation, on what evidence, and how the accused can appeal. The ruling does not say. Marked open in 5d, to be put to the operator.
  5. Also open: whether a decision that flips on recount is reversed, re-voted or only annotated.

Nothing in the sim implements this; it is not simulated. No code changed, nothing merged.

🤖 Generated with Claude Code

**Operator ruling, 2026-11-07 (#operator, Meta): option D.** > One account per person. If we determine a person to have broken the policy, their identities are consolidated and decisions they influenced are subject to a recount. Recorded in `SPEC.md` section 5d (commit 28d25ae, docs only; `REPORT.md` points to it). What it requires: 1. A durable per-decision vote ledger: voter identity, estate, weight at the time (with rule version), timestamp, ballot. Append-only. 2. An identity-consolidation operation with an audit receipt (who decided, evidence, identities merged, surviving account, time). It maps identities to a person; it does not edit the ledger. 3. A deterministic recount: same ledger, mapping and rule version always give the same tally; the consolidated person counts once, weighted by the combined purchase. 4. **Open question:** who determines a violation, on what evidence, and how the accused can appeal. The ruling does not say. Marked open in 5d, to be put to the operator. 5. Also open: whether a decision that flips on recount is reversed, re-voted or only annotated. Nothing in the sim implements this; it is not simulated. No code changed, nothing merged. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
GameCult/Bifrost!5
No description provided.