RUN  v2.1.9 <repo-root>

stdout | benchmarks/differential/harness.test.ts > differential benchmark: reactive side > runs chat/wide/selective/nogain and writes .results/reactive.json
┌─────────┬─────────────┬───────────┬────────────────┐
│ (index) │ workload    │ median_ms │ executedPerRun │
├─────────┼─────────────┼───────────┼────────────────┤
│ 0       │ 'chat'      │ '0.218'   │ 2              │
│ 1       │ 'wide'      │ '2.960'   │ 1              │
│ 2       │ 'selective' │ '1.847'   │ 10             │
│ 3       │ 'nogain'    │ '120.178' │ 1000           │
└─────────┴─────────────┴───────────┴────────────────┘

 ✓ |benchmarks| benchmarks/differential/harness.test.ts (1 test) 699ms
   ✓ differential benchmark: reactive side > runs chat/wide/selective/nogain and writes .results/reactive.json 698ms

 Test Files  1 passed (1)
      Tests  1 passed (1)
   Start at  15:54:00
   Duration  1.89s (transform 286ms, setup 0ms, collect 559ms, tests 699ms, environment 0ms, prepare 167ms)

[reactive] pnpm exec vitest run benchmarks/differential/harness.test.ts
[langgraph] uv run --directory <repo-root>/benchmarks/differential python langgraph_side.py

Differential benchmark (median of runs, same machine)
workload   reactive ms  langgraph ms     ratio
chat             0.218         1.147      5.3x
wide             2.960      3251.446   1098.3x
selective        1.847      3180.731   1721.8x
nogain         120.178      2750.486     22.9x

reactive executedPerRun: chat=2, wide=1, selective=10, nogain=1000
Outputs identical for every workload ✓