Don't struggle grasping text.
Instead, read the room.

Foreshock measures the human emotional and intent spectrum from the geometry of language alone. Understand your organization's morale, weigh the emotional impact of any content, and stop bad actors with a firewall your community curates that issues accessible receipts.

Purely semantic and does not use LLM classification
100+ Variable EV Dimension Keysspectrum-mapped and consistent Collects Zero PIIonly vectors and scores
● EV SPECTRUM 28 / 98 Sensors Above Kneedle
Valence
+0.31
Tension
RISING
Intent
CALM
Synthetic
3%
Runs where you already are
01 What Foreshock solves

Three blind spots, one instrument.

The same geometric engine reads inward at your organization and outward at the content reaching it.

Morale & its trajectory

Sample everyday communication and see underlying tension, fear, hedging or partisanism well before it surfaces as a visible crisis. You get the direction of travel, not just a snapshot.

Emotional impact of content

Score the estimated emotional impact of any user-generated or AI-generated content before it lands. Understand how it will be felt, on a graduated scale, across 75+ dimensions.

Transparent protection

Guard against external bad actors with full transparency, autonomy and agility. Inspect and edit your own triggers without a full retrain, and hand out receipts when you act.

02 The mandate

No black boxes. No LLM classification.

Open corpus
Inspect everything

The phrases that define every sensor and shunt are open and editable. Security through obscurity is not security.

Geometry only
Never reads content

Foreshock collects vectors and similarity scores. It does not comprehend, summarize or store your text.

No PII
Nothing regional, nothing personal

We know only which language and model were used. No PII, no regional data, no user profiles, ever.

On-prem ready
Run it yourself

Self-host inference on commodity hardware. The code and the math belong to you and your community.

03 Coming very soon

Vision, in the same vector space.

Foreshock uses Nomic Text and Nomic Vision by default, with 768-dimensional vectors and open alignment weights. A unified text-and-image space means one geometry for everything you publish. Got your own embedding model? Customize up to 1k+ dimensions, or trim to fewer, to suit your embedding needs. Contact us to discuss custom deployments in advance.

256768 default1k+3k
TEXT · nomic-1.5
VISION · nomic-v
UNIFIED SPACE
Be your community's own semantic hero

Safety that serves everyone that's designed by everyone, not an opaque model.

Start free for a community of 1 - 250. Connect a source and Foreshock starts scoring. Free tools to explore, learn, monitor, and respond. No pushy upsells.

We Really Mean Free Meet Aldous: The Model