AI-Lab Visual Showcase

EXP-007 Active

Direct visual inspection of generated simulation sandboxes, multi-agent effort telemetry, and Playwright verification.

Arm A: Vanilla Baseline (10,838 bytes, plain SVG flow) sandbox/exp007_vanilla/index.html ↗
Arm B: Governed Sub-Agents (17,719 bytes, animated pulses, Inter) sandbox/exp007_governed/index.html ↗

Live API Gateway Topology Sandbox

Interactive Canvas

Animated microservice flow with live chaos injection and real-time p50/p95/p99 sparklines.

Ingress Gateway Auth Service Healthy (2.1ms) Billing Service Healthy (4.8ms) Search Engine Healthy (6.2ms) Influx Sink Analytics DB
Latency (p50) 4.2ms
Tail Latency (p95) 12.8ms
Critical Tail (p99) 18.4ms
EXP-007 Cost Reduction
$0.0308 $0.0834 -63.05%
Vanilla Baseline:$0.08341
Governed Subagents:$0.03082
Thinking Tokens Rationed
1,410 2,150 -740 tok (-34.4%)

Reserving Max Effort for the chaos/traffic math while assigning Low Effort to Playwright runners prevented reasoning waste on terminal runs.

Cache Retention Shield
80.5% 5.9% +74.6%

Subagent sandboxes prevented terminal traces from polluting the parent coordinator context, sustaining 65,400 cached prompt tokens.