Between gatherings.
Community AI activity across Southeast Asia
Scroll to explore211 measured events · 11 countries
Before 10 September 2026, 00:00 UTC
Event totals and registration totals use separate, labelled scales.
One region.Different rhythms.
Our internal dataset, supplemented by verified public event reports, contains 235 completed events across eleven Southeast Asian countries. Registration measurements are available for 211 events, recording 22,445 registration instances. The snapshot ends before 10 September 2026 at 00:00 UTC.
The map changes as we change the measure. Malaysia and Singapore account for most sampled events. Vietnam becomes more prominent when we look at registrations. This is a comparison of the observed sample, with selective and uneven coverage across countries.
Registration instances can include the same person at different events. They are not a count of unique people, verified attendance or sustained AI use. Keep that distinction in view as we move through four better-observed country samples.
Sample definition and methodologyOct 2025–Sep 2026 · *September partial. Gaps preserve missing observations.
A calendar toreturn to.
Malaysia combines frequent small gatherings with occasional programmes of much greater scale. Its median event has 26 registrations across 100 measured events. That typical size sits alongside a total of 8,950 registration instances.
The contrast appears across months too. December records 1,773 registrations from seven events; August records 1,042 from twenty-two. More gatherings do not automatically produce a larger monthly registration total.
Our interpretation is that recurring sessions can provide somewhere to keep experimenting and asking for help. The calendar establishes repeated opportunities. It does not establish that the same participants return or that their experiments become sustained use.
Malaysia analysis and monthly evidenceDifferent rooms.Different scales.
Singapore combines focused exchange with larger opportunities to demonstrate work. Panels and roundtables have a median of 21 registrations. Demonstrations and showcases have a median of 150. Both formats appear within the same observed calendar.
Approval is recorded for at least 65 of 85 events. Access is often managed before a session begins. That is relevant context for interpreting the scale of participation; it does not establish who was admitted or why.
The opportunity is to connect useful exchange with a credible next step: a user, an operating problem or a supported trial. The data does not show whether people or projects actually move between these formats.
Singapore formats and access evidenceAt least 14 of 16 events
Jakarta or Tangerang
One community strongly influences this sample. Oct 2025–Sep 2026; *September partial. Zero means the format is absent in a labelled month.
Practical learning,repeated.
Indonesia’s observed programming includes workshops recurring between January and June. Recorded workshop registrations vary, but the series shows repeated practical sessions without the extreme programme peaks visible in some other samples.
This picture has a narrow geographic and organisational base. At least fourteen of sixteen events are in Jakarta or Tangerang, and a heavily represented community shapes much of the practical format mix.
The evidence supports recurrence within the observed programming. Extending similar opportunities to other places would test whether the approach transfers. It would also require attention to local language, access and the organisations able to support practical work.
Indonesia programming and coverage evidenceSmaller community calendars are less visible in the measurements.
Rounded concentration reported in the country analysis. Format labels overlap.
A few programmes.Substantial reach.
Vietnam records 5,640 registration instances. Its three largest events account for 83% of that total. The measured picture is therefore particularly sensitive to a small number of large programmes.
Some programmes carry several format labels. A workshop and a hackathon line can describe the same registrations. Adding those lines together would count overlapping activity more than once.
Smaller community calendars provide another layer of activity that is less visible in these measurements. Concentrated programme reach does not establish that recurring communities are weak. The practical question is how participants find further opportunities to learn and build after a large programme ends.
Vietnam concentration and coverage analysis| Country | Oct | Nov | Dec | Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep* |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MalaysiaMY | — | 7 | 7 | 2 | 1 | 8 | 8 | 8 | 15 | 22 | 22 | 8 |
| SingaporeSG | 1 | — | — | — | 1 | 2 | 4 | 19 | 9 | 18 | 21 | 10 |
| IndonesiaID | 1 | 1 | — | 2 | 1 | 4 | 3 | 1 | 1 | 2 | — | — |
| VietnamVN | — | — | 1 | 1 | — | 2 | 3 | — | — | 3 | 3 | — |
| ThailandTH | — | — | — | — | — | — | — | — | 2 | — | 1 | 2 |
| PhilippinesPH | — | — | — | — | — | — | — | — | — | — | 1 | — |
| BruneiBN | — | — | — | — | — | — | — | — | — | — | 1 | — |
| CambodiaKH | 1 | — | — | — | — | — | — | — | — | — | — | — |
| LaosLA | — | — | — | — | — | — | — | — | — | — | 1 | — |
| MyanmarMM | — | — | — | — | — | 1 | — | — | — | 1 | 1 | — |
| Timor-LesteTL | — | — | — | — | — | — | — | — | 1 | — | — | — |
October 2025–September 2026. *September partial.
— No sampled events; not verified absence of activity.
What keepsrecurring?
Bring the calendars together and another pattern becomes visible. Activity appears repeatedly in the better-observed samples, but its frequency and measured scale vary. A registration peak and a regular calendar are different signals.
These cells count events by their starting month. A blank cell means no sampled events, not verified absence of activity. September contains only the opening days before the cutoff. Changes in the sources captured can also change the apparent rhythm.
Recurring programming is observable here. Returning participants are not. Measuring whether people continue would require separate follow-up rather than an inference from a busy calendar.
Monthly activity grid and coverage notesAt least 82 events in the Klang Valley.
At least 14 in Jakarta or Tangerang.
8 in Ho Chi Minh City.
Documented event concentrations. Map shapes locate countries, not events or participants.
Coverage shapesthe picture.
All eleven countries have sampled activity. Six have no comparable registration measurements. Thailand has five sampled events and only three registration measurements. Wider geographic coverage makes more activity visible without making the samples equally representative.
The Philippines and Brunei contribute institutional building programmes; Cambodia contributes knowledge exchange; Laos contributes applied professional learning. Myanmar has repeated build programming, while Timor-Leste contributes an AI-literacy workshop. These observations deserve visibility without becoming national rankings.
Even the larger samples concentrate in metropolitan areas. Understanding what we cannot yet compare helps identify where better evidence would matter. Missing measurements remain missing; they never become evidence of an inactive community.
Country coverage and institutional contextWhat we observe
Gatherings
Recurring programming
Registration reach
What to support & test
Mentoring
Shared documentation
User testing
Continued use
Give the work somewhere
to continue.
After the roomempties.
Events make activity visible. The work between them is harder to observe: someone answering a technical question, documenting an experiment, finding a user or helping a team test what it built.
Our recommendation is to support that work alongside programmes. Mentoring, shared documentation and practical testing can give locally led communities more capacity to continue. Those connections are proposed routes to investigate, not participant journeys established by this dataset.
The next evidence should ask who returns, what keeps being built, what reaches users and what remains in use. For AISEA, the task is to help communities retain knowledge and practical support across gatherings, while preserving the local leadership that makes each community its own.
Evidence boundaries and full reportOur internal dataset, supplemented by verified public event reports.
Frozen before 10 September 2026, 00:00 UTC. Registrations are not verified attendance.