Path diversity
Resilience comes from independent, mission-capable alternatives—not from drawing more lines between components that share the same dependency.
A network is forming
Compose the path.
Test the failure.
Preserve authorized options.
KillWebs.com turns the shift from a fixed kill chain to a resilient, governed, increasingly autonomous network into an interactive synthetic learning environment. Inspect the nodes, remove a dependency, and watch the option space contract or recompose.
- 13
- abstract nodes
- 17
- possible relationships
- 7
- deterministic scenarios
- 0
- external runtime services
Current scenario
Baseline: several usable pathways
The preferred path is available and at least one alternate exists.
- Lesson
- A web is valuable only when alternative paths are compatible, trusted, and authorized—not merely drawn on the screen.
- Recovery
- Compare the highlighted route with the unused alternatives.
- Boundary
- Fictional nodes, illustrative state, no operational data.
Kill Web Explorer
One graph. Several possible chains.
Rotate the scene, select a node, degrade it, then ask the system to find another path that still passes compatibility, trust, and authority gates.
Canvas view unavailable
The complete node register and selected pathway remain available beside and below the visualization.
Topology snapshot
Count the options. Then inspect what they share.
Raw path count can overstate resilience when every route still depends on the same orchestration, policy, authority, timing, or software service.
- 8
- usable paths
- 3
- source options
- 2
- effect options
- 3
- shared dependencies
Accessible graph register
Every node outside the canvas
The visualization is optional. Keyboard users and non-Canvas clients can inspect the same state through these semantic controls.
Accessible relationship register
Every edge outside the canvas
A visible relationship can still be blocked by scenario state, compatibility, or authority. This table preserves the complete edge model for non-Canvas readers.
| From | To | Relationship | Current gate | Latency band | Failure domain |
|---|---|---|---|---|---|
| Wide-area sensor | Federated data service | data | Usable | seconds | space-path |
| Edge sensor | Edge processing | data | Usable | subsecond | local-path |
| Passive source | Federated data service | data | Usable | seconds | passive-path |
| Edge processing | Primary relay | transport | Usable | subsecond | mesh-path |
| Edge processing | Alternate relay | transport | Usable | seconds | alternate-path |
| Federated data service | Option composer | data | Usable | seconds | data-service |
| Primary relay | Option composer | transport | Usable | subsecond | mesh-path |
| Alternate relay | Option composer | transport | Usable | seconds | alternate-path |
| Option composer | Policy and trust gate | recommendation | Usable | seconds | policy-service |
| Policy and trust gate | Designated authority | authority | Usable | human-review | command-authority |
| Designated authority | Primary effect node | command | Usable | bounded | primary-effect |
| Designated authority | Alternate effect node | command | Usable | bounded | alternate-effect |
| Option composer | Alternate effect node | potential | Potential only | unknown | none |
| Federated data service | Primary effect node | potential | Potential only | unknown | none |
| Primary effect node | Assessment service | assessment | Usable | minutes | primary-assessment |
| Alternate effect node | Assessment service | assessment | Usable | minutes | alternate-assessment |
| Assessment service | Federated data service | feedback | Usable | minutes | feedback |
What the web changes
From protecting one route to preserving the outcome.
Later binding
Pair sensors, data services, authorities, and effect nodes when current availability, evidence, timing, and policy are known.
Explicit authority
A technically possible edge is not an authorized path. The model keeps compatibility, trust, and command decisions separate.
Graceful degradation
When the preferred route fails, continue a safe reduced function, expose the loss, and recompose only through validated alternatives.
Capability marketplace
How ACK turns options into comparable offers.
Explore a synthetic version of DARPA’s marketplace model: consumers request an outcome, providers advertise bounded capabilities, and a Virtual Liaison-style layer exposes constraints without assuming authorization.
Enter the ACK LabHuman authority
Where judgment actually lives.
Trace the roles of evidence review, policy, mission configuration, recommendation, authorization, supervision, abort, and accountability across the lifecycle.
Open the Authority LabAutonomous kill webs
See how distributed machine roles continue, reassign, quarantine, or hold.
Explore the general architecture across sensing, edge processing, state estimation, decision support, bounded non-force execution, assessment, human oversight, and institutional governance—without centering one weapon or platform.
Open the Autonomy LabAutonomy assurance
Inject compound failures before trusting autonomous behavior.
Test clocks, provenance, task ownership, signed updates, authority, human workload, intervention time, safe-state recovery, and reconnection as separate conditions—never as one reassuring score.
Open the Assurance LabAssurance traceability
See exactly why a claim is supported, qualified, suspended, or withdrawn.
Connect subclaims to dated evidence, assumptions, counterclaims, defeaters, verification, residual risks, accountable reviewers, and change events—without converting the graph into certification.
Open the Assurance Case LabMachine leadership
See how AI moves from tool to coordinator, allocator, and proxy leader.
Compare assistance, recommendation, coordination, delegated execution, gatekeeping, and governance across companies, public administration, municipalities, autonomous enterprises, and kill-web orchestration.
Open the Leadership ObservatoryLeadership accountability
Test whether machine-led decisions can actually be challenged, stopped, and remedied.
Keep notice, explanation, evidence access, correction, reconsideration, independent review, override, rollback, remedy, audit, vendor duties, and institutional responsibility separate.
Open the Accountability LabDecision provenance
Replay what the machine knew, changed, filtered, recommended, and caused.
Trace fictional source records through transformations, ranking, human review, execution, challenge, rollback, remedy, and later invalidating change—without confusing a complete-looking log with a justified decision.
Open the Audit Replay LabDelegation boundaries
See where machine coordination must pause, escalate, expire, or return control.
Inspect purpose, scope, tools, prohibitions, subdelegation, expiry, revocation, emergency exceptions, objective drift, human override, escalation deadlines, degraded-operation bounds, and accountability as separate controls.
Open the Delegation LabDelegation portfolio conflicts
Test what happens when several valid machine delegations collide.
Keep resource capacity, consumption authority, data purpose, objectives, institutional precedence, review windows, duplicate ownership, version alignment, revocation propagation, containment, and remedy separate.
Open the Portfolio Conflict LabAnticipatory intelligence
Forecast systems without turning foresight into pre-crime.
Build a synthetic probability, compare competing hypotheses, expose contrary evidence and shared provenance, then test calibration without predicting named people or creating authority.
Open the Anticipatory LabVulnerability surface
When a connected network is not a trustworthy network.
Examine integrity, timing, identity, data poisoning, supply-chain, electromagnetic, and common-mode risks at a defensive architecture level.
Study resilienceEvulgare machine answerability
Stop using software that blames the human.
Autonomous systems can make consequential decisions in milliseconds. When they fail, the nearest human should not automatically be forced to explain behavior they could not see, verify, understand, or control.
Let Evulgare’s machine intelligence answer for the machine. It takes responsibility for preserving and producing the technical account—what the system knew, what it did not know, which software and models were operating, what authority existed, what the human actually received, what action occurred, and where failure originated—so a person is not forced to invent an explanation for machine behavior they could not meaningfully review or control.
Sister-site bridge
The chain remains. The web changes how it is assembled.
Use KillChains.com to study a selected sequence and its interruption points. Use KillWebs.com to study the option space around that sequence—especially when nodes, links, data, or authority change.
Answer-ready summary
Direct answers
What is a kill web?
A kill web is a governed network of sensing, data, communications, command, support, and effect capabilities from which one or more authorized mission chains can be composed.
Read the supporting pageDoes a kill web replace the kill chain?
No. The web is the changing option space; a chain is the selected path used to sense, decide, authorize, act, and assess for a particular mission.
Read the supporting pageWhat makes a kill web autonomous?
Autonomy is distributed across sensing, state estimation, routing, recommendation, bounded task execution, assessment, and recovery. Force application, mission expansion, and governance remain separate high-consequence functions.
Read the supporting pageHow should autonomous kill webs be assured?
Test the combined human-machine system under compound faults, keep evidence, timing, trust, authority, intervention, containment, recovery, and reconciliation separate, and refuse to hide failures inside one readiness score.
Read the supporting pageHow is an autonomy assurance case kept current?
Each bounded claim is linked to evidence, assumptions, counterclaims, defeaters, verification, residual risk, review responsibility, and change events that can suspend or withdraw reliance.
Read the supporting pageHow is machine intelligence taking on leadership roles?
Machine intelligence exercises functional leadership when it controls attention, frames options, allocates resources, directs tasks, coordinates other agents, or executes bounded actions—even while formal legal authority remains human or institutional.
Read the supporting pageWhy should machine-speed systems carry their own technical explanation?
A human click should not become a liability transfer when a machine acted faster than the person could see, verify, understand, reject, delay, or control. Evulgare preserves the evidence needed to reconstruct the machine decision.
Read the supporting pageIs this an operational planning tool?
No. Every scenario, node, score, latency band, and outcome is fictional and designed only to teach architecture, evidence, resilience, and authority.
Read the supporting pageResearch corpus
60 preserved reports. One traceable memory system.
Every accepted uploaded report is stored under /docs/long-term-memory/reports, assigned a stable ID and SHA-256 digest, and routed through the project’s .uai memory without copying full report bodies into startup state.