Content analysis · 27 submissions
Every team deck and PDF submitted to the Singapore AI Architecture Hackathon in 2024, 2025 and 2026, read and tagged by theme. The shape of the entries changed more than the technology did.
Teams explicitly framing the build as an "agent" / "agentic" system
0 of 10 in 2024 · 9 of 10 in 2026
Teams working directly against BIM / Revit / IFC data
1 of 10 in 2024 · 6 of 10 in 2026
Teams quoting a quantified ROI, KPI or cost figure
0 of 10 in 2024 · 5 of 10 in 2026
What problem are they framed as solving
Primary category per submission, one per team. Hover a segment for the teams behind it.
How they're solving it — six signals, three years each
Share of that year's teams whose deck contains the signal, regardless of primary category. Hover a point for examples.
The thematic map — all 27 projects, clustered
A t-SNE projection of each project's 12-dimensional tag vector (category × six content signals), run live in your browser — perplexity 6, 600 iterations. Distance is thematic: projects tagged alike sit close together regardless of year; nothing here is spatial or geographic. Hover a point for details, click to jump to it below.
Who's building this — participant profiles, 2024–2026
Team rosters as disclosed in each deck's own team-member slide. Not every deck names its members, so the counts below are of disclosed rosters only — see the coverage note under each chart.
Team size, where disclosed
Each dot is one team, positioned by member count. Hover a dot for the team name.
Builders who came back
People (or intact squads) that appear on more than one year's roster — found by matching names across every disclosed team.
Firms that came back
Organisations named on more than one year's roster.
What roles look like, when a team actually lists them
Only 3 of 27 decks broke their roster down by discipline or title — too few to chart as a trend, so read these as examples, not a distribution.
Submission register
All 27 teams, tagged. Filter by year or theme.