legend
Ripples
What each weekly refresh actually changed
The chart refreshes itself once a week and says nothing when it succeeds. This is the record of what those runs actually found: every refresh that changed a figure, in the order it happened. Runs that found nothing new are not listed — silence is the normal state of a slow-moving archive.
The feed
2026-08-16 · 23:13 UTC
- nbim lost 1 record (70 → 69) — worth checking
2026-08-16 · 22:01 UTC
- Temasek announcements gained 2 records (46 → 48)
2026-08-13 · 23:30 UTC
- nbim-returns gained 1 record (28 → 29)
2026-08-07 · 23:30 UTC
- Temasek's 13F book gained 1 record (86 → 87)
Written by scripts/refresh.mjs after each run; only steps whose record count moved are listed. Machine-readable: RSS · changes.json.
How often this is checked
| Check | Cadence | Last run |
|---|---|---|
| SEC filings — a cheap index read for both funds | daily | 2026-09-12 |
| Everything — reports, Budgets, newsrooms, Norway | weekly, Mondays | 2026-09-06 |
The daily check reads only SEC's filing index; if nothing has been filed it does nothing at all. When something HAS been filed it triggers the full refresh, so a new 13F reaches this site within a day rather than waiting for Monday.
Current state
| Source | Records | Last run |
|---|---|---|
| official figures | 22 | unchanged |
| GIC's SEC filings | 141 | unchanged |
| Temasek's 13F book | 87 | unchanged |
| GIC announcements | 80 | unchanged |
| Temasek announcements | 48 | unchanged |
| nbim | 69 | unchanged |
| nbim-returns | 28 | unchanged |
| derived analysis | 29 | unchanged |
| the announcement↔filing join | 1 | unchanged |
| the prediction ledger | 5 | unchanged |
Run 2026-09-06 in 29s. A failed step never overwrites good data: the refresh restores the previous file and the site goes stale rather than wrong.