Different fields. Connected experience.
Change the lens to discover what connects the work. Choose a cluster or project to explore where your next challenge might fit.
49 projects 4 perspectivesBrowse all projectsReading the axes Position shows similarity. These directions are not business categories or skill scales.
Size reflects wow factor · perspective affects apparent size
This lens compares recorded fields with equal weight. Categories use normalized indicator vectors: exact shared values make projects closer, regardless of how many tags a field contains. Contribution also uses team-size ranges: log(1 + size), scaled to the current population; both endpoints share one field’s weight. Constant fields contribute no distance. Missing required values are excluded rather than filled with zero. Working pattern is a recorded category, not an estimate of availability or effort. Proximity is one minus the mean field dissimilarity. The projection uses the corresponding Euclidean distance, √(2 − 2 × proximity).
Only part of the original geometry fits on screen. Use the neighbour ranking for similarity in the full space; the projection is an overview. Across clusters highlights the selected project’s group and its closest project in each other group. The sidebar compares recorded tag frequencies between those groups, not inferred causes of clustering. Closest matches instead ranks up to five individual projects, regardless of cluster. Colours identify medoid clusters computed separately for the current data; their numbers are local to a map. In 3D, lines connect the selected project to the displayed matches. Network lines retain the nearest-neighbour graph and do not change with the sidebar mode. Connections do not establish collaboration or measured business outcomes.
Automatic grouping tests feasible counts from 2 to 10, with at least two projects per group and three deterministic PAM starts. It refits 12 subsets containing 80% of the projects, then picks the smallest count within one standard error of the best mean resampled silhouette. This is a simplicity heuristic, not a confidence interval. Stability is mean adjusted Rand agreement on retained projects; it is not accuracy. If no eligible separated partition is found, the map remains one group. Current separation: 0.292; stability: 0.688.
Faint convex envelopes surround each cluster in the displayed projection, with a small margin for the dots. Envelopes can overlap; they do not mark exclusive territories or a confidence interval. Turn off Clusters for an unobstructed view.
Bubble size uses the recorded wow-factor, a subjective portfolio rating from 1 to 10. Circle area scales with the rating; perspective also affects apparent size in 3D. Unscored projects use a small fallback marker. Visual sizing does not affect embeddings, cluster membership or neighbour rankings.
Source refreshed: 2026-11-18.Required fields must be recorded; this map is rebuilt locally when the snapshot is prepared. 0 projects excluded in this lens.
Bring your field into the picture
A familiar problem, a new industry, or a product that connects several worlds. Let’s explore what your project needs.
Discuss your project