Trust page
ArcSight passes when it proves impact — and passes when it refuses without proof. Refusal is a feature, not a failure. Outcomes, incomplete runs, and evidence links are published.
Trust path: Homepage → Validation → Repositories → 2026 report → Install · Architecture patterns
Verdict: ArcSight passes when it proves impact — and passes when it refuses without proof.
When structural evidence is incomplete, ArcSight stops — no synthetic confidence scores. That boundary is what makes merge recommendations trustworthy.
VS Code
0 dependency graph
Refused architectural conclusions. No fake confidence.
ArcSight produced impact analysis when evidence existed. ArcSight refused when evidence was incomplete.
Payload
Strongsrc/types fan-in 225
Focused investigation on the true platform contract.
Directus
Strong100-node cycle cluster
Reduced 231 findings into 3 architectural themes.
Nx
StrongWorkspace-level blast radius
Prevented a risky platform change before merge.
VS Code
Refused0 dependency graph
Refused architectural conclusions. No fake confidence.
10 additional repositories documented in the full record — same refusal-first criteria.
Incomplete evidence paths terminate with Analysis incomplete — no synthetic confidence scores.
Outcome: Refusal emits Reason, Evidence, Recommendation — never a invented risk score.
Evidence thresholds and refusal rules reviewed against production merge workflows.
Outcome: Complete analysis requires structural coverage; insufficient evidence always stops — no confidence inflation.
CLI commands and structural signals documented per release. impact, scan, and refusal paths are listed in the CLI reference.
Outcome: Command surface and structural signals are versioned — impact, refusal, and scan paths traceable per release.
Median install-to-first arc impact result under 3 minutes on a standard TypeScript monorepo (Node 18+, npm global install, no upload).
Outcome: Install → first meaningful arc impact output in under 3 minutes (median, internal runs).
Structured merge vs. hold scenarios run against real codebases. Engineers record decisions before and after arc impact output.
Outcome: Pre-output hold/merge intent recorded; post-output decision aligns with blast radius and break paths.
7 of 14 runs ended incomplete — evidence gaps, not false architectural reports. 0 false reports after graph gating.
Incomplete means: ArcSight refused or stopped short of a merge recommendation when structural coverage was insufficient. That is the intended trust boundary — not a silent wrong answer.
Each procedure links to the artifact used during v1 validation. Homepage stages reproduce the same schemas for live demo and frozen proof.
State of TypeScript Architecture 2026 →Next.js repository proof →Install ArcSight →← Back to homepage