Research log
Dated record of research dives that shape the platform. Each entry captures the question, what was found, the level of confidence, and what was decided as a result.
This is intentionally append-only. When facts change, add a new entry — don't rewrite old ones. The current-best summary of any topic lives in its own wiki page (e.g. Event-sources); this log is the audit trail behind those pages.
How to use this log
- One entry per dive. A "dive" is a focused question that produced a decision or a working inventory.
- Date it. Entries are ordered newest-first.
- Link out. If the dive produced or updated an inventory page (like Event-sources), link to it. Don't duplicate the table here.
- Capture caveats. Time-sensitive facts (API gate changes, deprecations) belong in the entry so future readers know when to re-verify.
- Record the decision. Even "decided to wait" is a decision worth logging.
2026-11-06 — Function view and national routes
Question: What does a parallel exploration of the same topic add to the G4 scan?
Method: Read a parallel working thread and checked its main claims against sources. The thread itself is model output and is not cited as a source.
Key findings:
- A view by function (participation, initiatives, neighbourhood networks, volunteering) shows the overlap better than a view by platform.
- Budget data is better than the scan suggested. Zorgzaam010 has a tender maximum of € 185,000 a year, and the Utrecht Initiatievenfonds has € 3 million plus € 900,000.
- VNG's collectivisation route is real, but digital public space is not among its first four topics.
- PubHubs and the Digital Commons EDIC were missing from this dossier.
Decision: Record the function view as a hypothesis in G4. Keep savings unclaimed until cost data proves them. Do not adopt a shared engine, automatic communities, or a new register as direction yet.
Caveats to re-verify: Rotterdam 2025 figures, DenkMee user count, Gebiedonline community and user counts, the G4 chair, the governance precedents, and PubHubs' current status.
Source pages updated: G4.
2026-11-05 — G4 scan: Rotterdam, Den Haag, Utrecht and their regions
Question: What does the digital public space look like outside Amsterdam, and does the atlas hold up when it moves to the other G4 cities?
Method: Four parallel research agents: one per city region, and one for the national layer (OpenStad, Gebiedonline, vendors, G4 and VNG channels, area models). Endpoints were requested directly. Conflicts between the agents and all corrections to existing atlas claims were retested by hand against the live URLs.
Key findings:
- No two cities share a citywide platform: Go Vocal (Utrecht), Mendix (Rotterdam), OpenStad plus the municipal website (Den Haag), Gebiedonline plus OpenStad (Amsterdam).
- The neighbourhood layer differs in kind: Gebiedonline cooperatives (Utrecht), a municipal register of separate sites (Den Haag), elected councils with no resident-built network found (Rotterdam).
- The CBS Wijk- en Buurtkaart through PDOK returns neighbourhoods for all four municipality codes (CC BY 4.0).
- No G4 body on digital participation found. The nearest is the G4 CIO procurement agreement.
- The claim that Utrecht has a better engagement model is unsupported as written. The source is a personal observation in a meeting.
Decision: Keep Amsterdam as the deep layer and add the other cities as a lighter inventory at /atlas/g4. Do not add them to PLATFORMS, because the coverage grid and the open-field count assume Amsterdam profiles. Full profiles come after the operators have had a chance to correct the list.
Corrections to earlier entries:
/rsson Gebiedonline networks delivers per-network items, not only on Hallo IJburg. The channel title and link host stay wrong, and items have nopubDate. gebiedonline.nl returned 1 item and wijamsterdam.nl 0.- The OpenStad figure of about fifty organisations was attributed to a VNG page that does not contain it. VNG now states 43 public organisations and 500+ projects.
Caveats to re-verify:
- Rotterdam coverage is the weakest: many guessed addresses timed out, which proves nothing.
- Go Vocal's
/web_api/v1is public but undocumented. Reuse status is unknown. - Feed counts are a seven-day calendar window, not a publication count.
- Items based on search summaries carry
to-validate.
Open follow-ups: Utrecht's 2025 market consultation outcome, the announced Den Haag platform, neighbourhood sites in Leiden, Delft, Zoetermeer, and Westland, and a counterpart per city.
Source pages updated: G4 (new), Atlas (correction added), /atlas/interop, /patterns.
2026-06-02 — Map & geometry sources for Amsterdam
Question: What cartographic and geometry sources can render the Uitwijken map — base tiles, administrative boundaries (stadsdeel/wijk/buurt), address/building geometry, and street-level detail? This is the map counterpart to the event-pins dive: Open-data named the data types as principles, but the base cartography had never been inventoried at API level.
Method: Three parallel research agents — national PDOK/Kadaster services, Dutch basisregistraties (CBS/BAG/BGT/NWB), and Amsterdam's own municipal geo APIs. Endpoints live-probed against official docs.
Key findings:
- The map foundation is fully buildable on open data: PDOK (BRT-A base tiles, BAG, BGT, Locatieserver geocoder) + data.amsterdam.nl (gebieden, basiskaart, meldingen), free and no-auth.
- Use Amsterdam's own
gebiedenindeling for the platform's scales — it does not map 1:1 to CBS (carries a separatecbsCodefield). CBS Wijk- en Buurtkaart is the source for national comparability + statistics. - Licensing is mixed: BGT/NWB are CC-0, BRT-A/CBS are CC-BY (attribution needed), Amsterdam basiskaart has bespoke municipal terms.
- Two distinct BAG APIs: PDOK OGC API (no key, bulk/tiles) vs. Kadaster Individuele Bevragingen (free key, single lookups, rate-limited) — don't conflate.
Decision: Adopt the stack captured in Map-sources. Build the proof-of-concept map foundation on PDOK + data.amsterdam.nl with no commercial cartography dependency. Consider the base-cartography question closed; the civic-overlay layer (permits, plans, budgets) remains open.
Caveats to re-verify:
- API keys are rolling toward mandatory on both PDOK BAG (Individuele Bevragingen) and data.amsterdam.nl, with no firm deadline — biggest forward-looking risk.
- Exact BAG/BGT collection identifiers — confirm against live
/collectionsJSON before coding field names. - Amsterdam
gebiedenshows "Licentie: -" while behaving as open/CC-BY — confirm attribution terms.
Open follow-ups: Municipal permits/plans datasets, parkeervakken/bomen and other thematic map layers, BGT vector-tile render budgets at buurt scale, buurtbudget/Stadspas geo channels.
Source pages updated: Map-sources (new), Open-data (cross-link added).
2026-06-02 — Event data sources for Amsterdam
Question: Which sources of geotagged event data exist for Amsterdam and Noord-Holland, and which expose APIs / feeds with title, date, location, and source URL?
Method: Multi-agent deep-research workflow — 6 search angles, 27 sources fetched, 25 claims adversarially verified (24 confirmed, 1 refuted).
Key findings:
- Gemeente Amsterdam's Datapunt Evenementen dataset is the strongest single source: public, no-auth, CC-BY 4.0, REST + WFS + MVT.
- Most commercial event APIs (Luma, Meetup, DICE, Eventbrite) are partner-gated, paid, or owner-scoped.
- Ticketmaster Discovery API v2 is the most accessible commercial source — free key, 5k calls/day.
- Facebook Events API has been effectively closed since COVID; the canonical community scraper was archived in 2021.
- Major Amsterdam cultural venues (Melkweg, Paradiso, Pakhuis de Zwijger) publish no API/RSS/iCal — HTML scraping (ideally via embedded Schema.org JSON-LD) is the path.
Decision: Adopt the ingestion priority captured in Event-sources. Ship the "Events this week" map layer on open data + JSON-LD scraping with no commercial dependency. Defer commercial APIs until a business case justifies the cost.
Caveats to re-verify:
- Datapunt is rolling out a mandatory API key — code should be ready to send it.
- Meetup tightened its gate in Feb 2025 (Pro tier required for new OAuth consumers); older blog posts are misleading.
- Venue "no API" claims rest on absence-of-evidence — worth a manual check.
Open follow-ups: Uitagenda / Uit in Nederland, Songkick / Bandsintown / RA paid tiers, civic event channels (Decidim, buurtbudget, Stadspas), Museumnacht / UvA / VU / warenmarkten feeds.
Source pages updated: Event-sources (new), Open-data (cross-link added).
Tags
#year/2026 #city/amsterdam #process/research