SYNTHESIZEA SEMANTIC SYNTHESIZERHow it works

A SEMANTIC INSTRUMENT

Find the idea
between ideas.

A tool for exploring what exists between two ideas.

Enter any two pieces of text — a concept, a conversation, a poem, a research excerpt, or anything else.
Blend their meanings. Explore what lies near—or far from—the intersection.

01 / SOURCE MATERIALAnything expressible as text.

a poem · a research excerpt · a conversation
a brainstorm · anything else

0 / 5,000

a poem · a research excerpt · a conversation
a brainstorm · anything else

0 / 5,000
50 / 50
SOURCE ASOURCE B
0.00
NEARFAR
VOCABULARY

Familiar, broadly understood vocabulary.

18
1 KEYWORD100 KEYWORDS

Two sources, a spectrum of meaning.

Add text to both sources to synthesize.

02 / SEMANTIC FIELD

The language around your synthesis.

A space for what appears.

Your synthesis will resolve here.
Two sources. One semantic space.

RETRIEVED LANGUAGE. OPEN TO INTERPRETATION.

How it works

Synthesize treats meaning as something you can position, calculate, and explore.

  1. Source A and Source B are converted into semantic embeddings: numerical representations of relationships in language.
  2. Balance controls how much each source contributes. At 50 / 50, both influence the synthesis equally.
  3. Novelty changes how the two embeddings are combined. NEAR begins with their direct blend. Moving toward FAR increasingly emphasizes the difference between them, opening a more remote semantic neighborhood. It is an experimental control, not a calibrated measure of distance.
  4. The resulting vector is compared against the selected noun index. COMMON searches familiar vocabulary; NICHE opens the larger, stranger vocabulary.
  5. The nearest nouns become the semantic field. Classical multidimensional scaling (MDS) projects the relationships between their embeddings into two dimensions, so related words tend to appear near one another. The cyan points preserve those projected positions; labels shift only to remain legible. Type size reflects similarity to the synthesis.
  6. The field is a local projection, not a literal map of semantic space. Its orientation can change when the result set changes.
  7. The words are retrieved from the index, not written as an answer by an LLM. The field is something to explore, not an authoritative interpretation of your sources.

Synthesize explores one computational representation of semantic relationships, rather than an objective model of human meaning.