this.me / Benchmarks

Benchmarks 1–11 · Run #001

1 — Algorithmic Scaling 341.71ms

Wave k stays at 2 regardless of N. The engine touches only actual dependents — not the full dataset.

N (nodes)	time_ms	wave k	result
10	0.1094	2	100
100	0.0746	2	100
1,000	0.0841	2	100
5,000	0.0542	2	100

2 — Extended Scaling 1193.38ms

Flat to 10,000 nodes. k=2 throughout.

N	time_ms	wave k
10	0.1080	2
100	0.0534	2
500	0.0534	2
1,000	0.1051	2
2,500	0.0686	2
5,000	0.0532	2
7,500	0.0518	2
10,000	0.0521	2

3 — Incremental Processing 754.39ms

N	time_ms	wave k	status
10	0.1055	2	OK
100	0.0533	2	OK
1,000	0.0728	2	OK
5,000	0.0559	2	OK
10,000	0.0789	2	OK

4 — Multi-Dataset Stress Lab 386.57ms

Deep nesting (500 levels), wide broadcast (1,000 nodes), financial dataset (5,000 tx) — all k=2.

Dataset	Latency	wave k	Status
DEEP_NESTING (500 levels)	13.2138ms	2	✅ Reactive
WIDE_BROADCAST (1,000 nodes)	0.0864ms	2	✅ Reactive
FINANCIAL_DATASET (5,000 tx)	0.0789ms	2	✅ Reactive

5 — Throughput Under Sustained Mutation 191.92ms

2,000 consecutive mutations. p95 drift: −30.16% — gets faster, not slower.

p50	p95	p99	max
0.007ms	0.011ms	0.018ms	0.331ms

6 — Fan-Out Sensitivity 277.66ms

Latency decreases as fanout grows — larger datasets amortize index lookup. k stays 2.

fanout	k	p50_ms	p95_ms	p99_ms	max_ms
10	2	0.0124	0.0189	0.0253	0.0934
100	2	0.0082	0.0160	0.0201	0.0230
500	2	0.0074	0.0105	0.0161	0.0184
1,000	2	0.0062	0.0078	0.0102	0.0193
2,500	2	0.0061	0.0083	0.0120	0.0134
5,000	2	0.0057	0.0067	0.0089	0.0105

7 — Cold vs Warm Runtime 239.26ms

Cold start is sub-millisecond and absorbed after the first mutation.

nodes	cold_ms	warm_ms	steady_avg_ms
100	0.1735	0.0865	0.0142
1,000	0.0097	0.0113	0.0081
5,000	0.0149	0.0113	0.0067

8 — Explain Overhead 223.64ms

explain() adds ~0.007ms at p95. Full derivation traces at negligible cost.

mode	p50_ms	p95_ms	p99_ms
baseline	0.0077	0.0122	0.0199
with_explain	0.0129	0.0189	0.0250

9 — Secret-Scope Performance Impact 631.34ms

~27× overhead at p95 — expected AES-GCM cost. Keep hot-path reads on public branches.

scope	p50_ms	p95_ms	p99_ms
public	0.0109	0.0206	0.0594
secret	0.4951	0.5592	0.7615

10 — Push vs Pull (Eager/Lazy) 474.46ms

Both modes converge at scale. Lazy: lower mutation cost. Eager: lower first-read cost.

mode	fanout	k	mut_p95	read_p95
eager	10	2	0.0132	0.0114
eager	5,000	2	0.0036	0.0030
lazy	10	2	0.0027	0.0047
lazy	5,000	2	0.0037	0.0039

11 — Secret Push vs Pull 982.44ms

Secret read cost grows with node count — each cold read re-derives key material. Writes are cheaper than reads in encrypted branches.

plane	nodes	mut_p95	read_p95	read slowdown
public	100	0.0136	0.0183	—
secret	100	0.0409	0.3675	20×
public	300	0.0108	0.0123	—
secret	300	0.0313	0.2864	23×
public	600	0.0045	0.0041	—
secret	600	0.0344	0.5805	142×

Summary all passed

Benchmark	What it proves	Time
1	O(k) flat recompute (10 → 5,000 nodes)	345.98ms
2	Flat scaling to 10,000 nodes	1179.37ms
3	Incremental processing stability	753.01ms
4	Multi-shape stress (deep / wide / financial)	383.35ms
5	Sustained throughput, no p95 drift	188.64ms
6	Fan-out sensitivity (latency improves at scale)	275.75ms
7	Cold vs warm startup cost	234.70ms
8	explain() overhead (~0.007ms at p95)	222.43ms
9	Secret scope cost (~27× vs public at p95)	632.75ms
10	Eager vs lazy push/pull tradeoff	469.08ms
11	Secret push vs pull at scale	934.32ms