pulse-base-v1

An embedding model for time. Not timestamps -- time.

PULSE encodes moments as 128-dimensional vectors that capture not just when something happens, but what that time means -- urgency, circadian phase, behavioral context, and the felt sense of time.

What makes this different

Every temporal encoding in AI today treats time as a coordinate. PULSE treats it as an experience:

Capability	Time2Vec	RoPE	Neural ODE	Calendar	PULSE
Periodicity	✅	✅	❌	✅	✅
Calendar-aware	❌	❌	❌	✅	✅
Context-dependent	❌	❌	✅	❌	✅
Urgency/deadlines	❌	❌	❌	❌	✅
Circadian phase	❌	❌	❌	❌	✅
Experiential time	❌	❌	❌	❌	✅

Architecture

128D embedding from seven fused signal layers:

PULSE(t, context) = normalize(concat[
    log_time(t)         * 1.0,   # 8D  - Weber's Law compression
    oscillators(t)      * 0.5,   # 32D - Multi-frequency sinusoids
    circadian(t)        * 1.5,   # 8D  - 24h + 90min biological clock
    calendar(t)         * 0.6,   # 24D - Day/month/season/holiday
    urgency(t,deadline) * 4.0,   # 8D  - Hyperbolic deadline proximity
    temporal_state(h)   * 2.0,   # 32D - Continuous-time event history
    prediction_error(t) * 3.0,   # 16D - Temporal surprise
])

Usage

from pulse_temporal import PulseEncoder

pulse = PulseEncoder()

# Same hour, completely different moments
monday_crunch = pulse.encode("2026-04-13T14:00:00", context={
    "deadline": "2026-04-13T17:00:00",
    "events_today": 6,
    "sleep_hours": 5,
})

saturday_chill = pulse.encode("2026-04-11T14:00:00", context={
    "deadline": None,
    "events_today": 0,
    "sleep_hours": 9,
})

# These are FAR apart in PULSE space despite similar timestamps
pulse.similarity(monday_crunch, saturday_chill)  # ~0.72

# These cluster together -- "crunch time before deadline"
wednesday_crunch = pulse.encode("2026-04-15T10:00:00", context={
    "deadline": "2026-04-15T12:00:00",
    "events_today": 4,
})
pulse.similarity(monday_crunch, wednesday_crunch)  # ~0.78

Install

pip install pulse-temporal

Version

v0.1.0 -- formula-based encoder (no trained weights). All seven layers use principled formulas from neuroscience and behavioral economics. Trained model (v0.3) will use contrastive learning on human activity data.

Citation

@software{pulse_temporal,
  title={pulse-temporal: Experiential Time Embeddings for AI},
  author={Morales, Lalo Adrian},
  year={2026},
  url={https://github.com/lalomorales22/pulse-temporal}
}