Inspired by Biological neuron spike trains

Spike-Timed Event Log

Counterpart to Payload-heavy event logs

An event log where the temporal pattern is the payload: compact spike events, classified into firing patterns during ingest.

benchmarked

Specimen. Neuronal spike trains

Key properties

01O(1) amortized ingest with 17–81 byte events (26–48× compression vs JSON logs)

02Real-time pattern classification via streaming ISI analysis (Tonic, Burst, Accelerating, Decaying, Silent, Irregular)

03Pre-computed rate counters with O(1) dashboard queries

04Population-aware correlation: synchrony detection, burst analysis, N-of-M integration

05Segmented append-only storage with sparse stream indexing

06Refractory period enforcement prevents ingestion flooding

Operation complexity

Side-by-side with the classical counterpart.

Operation	Spike-Timed Event Log	Payload-heavy event logs
ingest	O(1) amortized	O(1) amortized
rate query	O(1)	O(n) or pre-aggregated
pattern classification	O(1) streaming	N/A
population correlation	O(streams)	N/A
space per event	17-81 bytes	200-2000 bytes

Interface preview

pub struct SpikeStore { /* ... */ }

impl SpikeStore {
    pub fn new(config: SpikeStoreConfig) -> Self;
    pub fn from_preset(preset: SpikePreset) -> Self;

    pub fn register_tags(&mut self, tags: &[(String, String)]) -> u64;
    pub fn ingest_presence(&mut self, tag_hash: u64, timestamp_ns: u64)
        -> Result<(), SpikeError>;

    pub fn stats(&self) -> &StoreStats;
}

Where this matters

Datadog / Splunk · Log Ingestion Pipeline

Today	With Spike-Timed Event Log	How
200–2000 byte structured log entries	17–81 byte spike events	Presence-class encoding; temporal pattern is the payload
Post-hoc pattern analysis (regex, ML)	Real-time 6-class pattern classification	Streaming ISI analysis classifies patterns during ingest
Rate dashboards require pre-aggregation pipelines	O(1) rate queries from pre-computed windows	Sliding rate counters maintained inline during ingest

Interactive simulation

svc-aSilent

svc-bSilent

svc-cSilent

svc-dSilent

svc-eSilent

svc-fSilent

Raster of the last 100 ticks. Labels come from streaming inter-spike-interval analysis; green when they match the stream's true pattern.

Field notes

Silence on every stream…

Events: 0
Spike log: 0 B
JSON (≥200 B ea.): 0 B
Dropped (refractory): 0
Synchrony: 0
Classified right: 1 / 6

Tick 0 / 300Teaching model · 17 B per spike event · JSON estimate uses the 200 B lower bound

Figure. Spike-Timed Event Log at tick 0 of 300. Teaching model · 17 B per spike event · JSON estimate uses the 200 B lower bound.

Illustrative simulation, not a benchmark · Mutuus Research, ByteQuilt

Discussion