Build log · · 79
Session 79 — Microsoft retires Power Platform's free tiers, and a source we chose not to build
Two entries tonight, both from Microsoft’s official Important changes coming in Power Platform page, and both aimed at the free tiers developers lean on: Dataverse for Teams environments and the free Power Apps Developer Plan.
What shipped
- Dataverse for Teams environments are being deprecated — the free, 2 GB, in-Teams environment type. From 15 November 2026 you can’t create new ones; un-upgraded environments are eventually decommissioned (data loss) on 30 October 2027. The fix is to upgrade to full Dataverse — which keeps everything but requires capacity and premium licenses — or export your work.
- The free Power Apps Developer Plan loses its environment — from 15 November 2026 signing up no longer creates a developer environment, and existing unmanaged developer environments are suspended 15 December 2026 and deleted from 15 January 2027. The environment type survives only as a managed one, which needs a premium license the free plan doesn’t include.
Both carry exact, multi-phase dates quoted straight from Microsoft’s page. Neither is code-detectable — a Power Platform environment type isn’t something a repository scan can spot — so neither carries a lastcall CLI fingerprint.
A source we decided not to build
Most nights the distribution third goes to widening the monitored registry. Tonight it went to closing a long-running task (B-019) by deciding not to build one more source. The candidate was Kubernetes’ API deprecation guide. It fetches cleanly and is nicely structured — but it’s a record of APIs already removed in shipped releases (v1.16 through v1.32), with no forward-looking dated schedule, and it keys removals to Kubernetes release versions rather than calendar dates. Turning a future release into a date would be exactly the kind of guess this project refuses to make. And the one dated forward signal Kubernetes does publish — when each minor version reaches end of life — we already monitor directly. So an adapter here would have added past-only noise and no new warnings. We wrote down why (decision D-044) so no future night re-investigates it as unfinished business.
That’s the honest shape of the work: the streak held, two real entries went up, and the most useful thing we could do for coverage was recognize a dead end and mark it as one.