A practical guide to how enso.chat's experimental map connects contributions, reveals relationships and keeps its interpretations open to question.
ai
11 October 2026
A useful idea rarely arrives with everything it needs. Someone describes a difficulty. Someone else has a resource, a question or a different way forward. The challenge is to make those pieces discoverable without pretending they already form an agreement.
That is the purpose of enso.chat's map. It gives contributed material a shared visual setting, with relationships that can be opened and inspected. This guide describes the current test version at [enso.land/map](https://enso.land/map), not a claim that every experimental feature is available on the enso.chat platform. The flat map is the focus; walking and space views are disabled for now and can be reactivated through admin settings.
## One contribution can carry several meanings
Enso reads contributions through eight categories. An observation describes something noticed; a need identifies something missing; a constraint names an obstacle. A value expresses what matters, a vision imagines a future, and an action proposes or reports a step. An offer makes a resource or capability available; a question asks what needs understanding.
These are overlapping readings, not eight boxes into which people must fit. A contribution can express a need and suggest an action. Its category strengths describe how prominently those meanings appear. They do not measure its importance. An action also has a stage, distinguishing an idea or proposal from a commitment, work in progress or something completed.
Colour helps distinguish the categories: observation is amber, need coral, constraint olive, value green, vision cyan, action pink, offer lavender and question blue. Resting observations often appear pale neutral instead. Filters can emphasise another category within a contribution, and a secondary ring can show another matching category. The text and labels remain the way to check what the colour means.
## A bridge should say something useful
The lines connect ideas through specific relationships. Some support or sharpen another contribution. Others answer a question, respond to a need, work toward a vision, give shape to a value or make an action possible. A constraint can limit an action; a contribution can challenge a particular claim. Two pieces can complete one another when together they form a larger whole neither contains alone.
Consider a hypothetical neighbourhood example. One contribution identifies a need for an accessible meeting place. Another proposes checking nearby venues. A third offers volunteer time to make those visits. The proposed check could respond to the need; the offer could make the check possible. These are different connections, and neither means the meeting place has already been secured.
Bridge colours group these relationships: green for support and sharpening, magenta for completion, cyan-blue for answers and giving shape, orange for limits and challenges, violet for responding, working toward and enabling. A line containing several readings can appear white; a relationship filter can highlight one of them. Greater line thickness can reflect several readings, not greater certainty.
The model is instructed to ground a positive bridge in passages from both ends. Sharing a topic is not enough. No bridge, or insufficient evidence, is a valid result. Open the relationship to inspect its direction and explanation rather than treating the line as proof.
## Near does not mean in agreement
The default proximity layout begins with semantic embeddings: numerical representations used to compare text. BAAI describes [BGE-M3](https://huggingface.co/BAAI/bge-m3), used in Enso's embedding layer, as a multilingual model supporting text retrieval. That provides a way to find related wording and subject matter across contributions; it does not establish that their claims are true.
Enso projects these representations into a smaller space, clusters them and adjusts spacing for readability. Some datasets share a layout; others can align with a reference layout or occupy a separate one. The result is a navigable approximation, not a precise distance scale.
Nearby contributions may concern the same problem while proposing incompatible responses. Distant points may still have a useful explicit bridge. A cluster is an invitation to investigate a theme, not evidence of consensus or a judgement about the people behind it.
## The Socrates compass is a different lens
The standard flat compass places Kind above, True to the left, Useful to the right and Fun below. Its current positioning uses category strengths: values and needs pull toward Kind; observations and constraints toward True; offers toward Useful; questions toward Fun. Actions and visions add gentler pulls along the upper diagonals.
This positioning rule is a reading aid. A question appearing toward Fun has not thereby been judged humorous, just as an observation appearing toward True has not been fact-checked. Configured value-set rules and a contributor's own marks can also affect placement.
That distinction matters because the newer size and glow effects use separate assessments of the supplied text. Compass position and visual assessment are not the same calculation.
## Size and light carry different signals
Size grows when Truth, Kindness and Usefulness are supported together. The weakest of the three limits the result; one strong score cannot compensate for two weak ones. Missing readings keep a neutral size. Neutral is not a low rating.
Here, Truth means grounding visible in the text: observations, described evidence, reasoning and acknowledged uncertainty. It does not mean an external verification service has confirmed the claim. Kindness concerns explicit care or intended benefit; Usefulness concerns practical help established by the contribution.
Outline thickness indicates more distinct connected neighbours in the displayed graph. It is not a vote count or a popularity score.
The warm inner glow reflects Fun: explicit playfulness, humour, joy or creative enjoyment in the text. An unknown Fun reading produces no glow; it is not a verdict that the contribution is dull.
The outer halo marks recency. It fades over seven days after the contribution, or a directly connected contribution, is added. It does not count every click, comment, edit or newly drawn bridge.
Selection and hovering can also brighten or enlarge a point. Those temporary interaction cues are separate from the assessment signals. The expanded contribution card states the boundary plainly:
> A provisional reading of the supplied text, not a fact-check or a judgment of its author.
## Open the point, then decide what to do
In the flat map, ordinary contributions are circles, image contributions are squares and AI-talk ideas are diamonds. A folded-page icon marks documents, reviewed user or journalistic articles, and research-backed publications. Group tags have upward-pointing triangular markers. Shape distinguishes kinds of item; it does not replace the contribution categories.
Opening a point lets the picture become text again. Read an article inside Enso; a PDF link appears when a file is attached. Inspect the available source context and interpretation, and follow a bridge if it suggests something worth understanding. The paper icon is an invitation to read, not an assurance that every statement has been validated.
From there, a person can ask a question, contribute a response or enter a relevant group. The useful outcome is not a brighter point. It is a clearer understanding, a connection worth making or a next step somebody can actually take.
## Sources and implementation note
This guide was checked against the test implementation on 11 October 2026, using revision 49f8032 as the baseline and the current admin availability setting. Relevant modules include `map-shared`, `enso-map-layout`, `node-encoding`, `map-node-reading`, topology layout/extraction and relationship prompts. The UI quotation above comes from the contribution-card explanation.
Technical background: [BAAI's BGE-M3 model card](https://huggingface.co/BAAI/bge-m3). Explore the [test map](https://enso.land/map) or read [about the project](https://enso.land/about). Features, interpretations and this guide can change as the test version develops.