Engine layer
Declare a target. Get a calibrated forecast of what's next.
This is the wedge. MemMesh doesn't just remember — it forecasts. Declare any target (an event, a number, a time, an anomaly) and the engine predicts it from a subject's history, with a confidence score checked against what actually happened and recalibrated over time. When it says 80%, it means it.
lattice.predict (event · numeric · time · anomaly)
What you get
Forecast any target
Event, number, time-to-event, or anomaly — one general predict() surface.
Calibrated confidence
Every forecast is scored against reality and recalibrated, so the number is honest.
Abstains when unsure
Thin signal yields a hedge or a pass, not fabricated certainty.
How it works
The loop, three calls.
- STEP 01
Learn from history
Predictions build on the subject's accumulated memory.
- STEP 02
Declare a target
Ask for any outcome; the engine forecasts it with a confidence score.
- STEP 03
Record & recalibrate
Feed the outcome back; the next forecast gets sharper.
lattice.predict (event→numeric→time→anomaly)Features
Predict any target
Event, numeric, time-to-event, or anomaly — one general predict() surface, not a fixed menu.
Calibrated confidence
Every forecast is scored against reality and recalibrated, so the number is honest.
Abstains when unsure
Thin signal? The engine hedges or abstains instead of fabricating certainty.
FAQ
Questions, answered.
What can it predict?+
Any declared target — an event, a numeric value, a time-to-event, or an anomaly — from a subject's history.
How honest is the confidence?+
Calibrated against actual outcomes: across the times it says 80%, the event should happen about 8 in 10. When the signal is thin, it abstains.
More engine layers
Memory
Recall that compounds.
Cross-session persistent memory with hybrid search and a knowledge graph. The engine decides what's worth keeping, so context carries between conversations without bloating your prompts.
memory.observe · memory.search · memory.reflect
Behaviors
From patterns to next-best-action.
The engine mines behavior into emergent patterns and closes the loop: predict, let the agent act, record the outcome, and watch the patterns recalibrate. Decisions and outcomes are first-class.
lattice.mine · learning.recordDecision / recordOutcome
Foresight, not just recall.
Declare any target and the engine predicts it from a subject's history — calibrated, provenanced, and abstaining when the signal is thin. This is the layer mem0 and MemoryLake have no answer for.