Microsoft refresh group 1 of 5 — October 28, 2026¶
The user requested completion of every remaining group with a commit/push and remaining-work update after each. This run froze 48 eligible guides in five risk-ordered groups of up to ten. This first group covers ten guides and 623 current detailed objectives; 38 guides in four groups remain after this group. Recently reviewed guides and retired archives were not silently added back to the queue.
Three read-only role researchers supplied exam, selected technical and learning-resource evidence. The coordinator owns shared edits. The operations ledger includes exact final objective/guide/span bindings, opened source observations, old-to-new snapshot hashes, repair scope and limits. Complete-guide reading is recorded for the exam researcher; the other workers describe their selected semantic scope. This is AI-assisted content review, not human review, a full independent semantic audit or live service execution.
| Guide | Current objectives | Source gate | Link evidence |
|---|---|---|---|
| DP-300 | 94 | Passed; human review pending | 77/80 reachable; 3 automation-blocked; 0 missing/error |
| DP-420 | 56 | Passed; human review pending | 89/92 reachable; 3 automation-blocked; 0 missing/error |
| DP-750 | 77 | Passed; human review pending | 70/74 reachable; 4 automation-blocked; 0 missing/error |
| DP-600 | 41 | Passed; human review pending | 41/44 reachable; 3 automation-blocked; 0 missing/error |
| AI-500 | 55 | Passed; human review pending | 40/42 reachable; 2 automation-blocked; 0 missing/error |
| MS-102 | 54 | Blocked by retained source conflict | 79/82 reachable; 3 automation-blocked; 0 missing/error |
| AB-650 | 58 | Blocked by retained source conflict | 54/56 reachable; 2 automation-blocked; 0 missing/error |
| DP-700 | 54 | Passed; human review pending | 50/53 reachable; 3 automation-blocked; 0 missing/error |
| PL-400 | 88 | Passed; human review pending | 56/57 reachable; 1 automation-blocked; 0 missing/error |
| AB-210 | 46 | Blocked by retained source conflict | 44/46 reachable; 2 automation-blocked; 0 missing/error |
Objective acceptance and history¶
Six current English outlines are accepted after teaching reconciliation: DP-300 (October 27), DP-420 (October 6), DP-600/DP-700/DP-750 (October 19), and the currently published PL-400 outline (October 16). Every prior objective byte is archived under its original hash; historical review/audit path migrations change only the pointer to those exact bytes. Dates, hashes, checks, counts, findings and verdicts remain unchanged. Four unchanged outlines are retained.
The initial matrix maps 687 historical objectives; the final matrix maps 623 current objectives. DP-420 changes from 132 July bullets to 56 current bullets, and PL-400 from 76 historical bullets to 88 current bullets. These totals are not interchangeable. Eighteen implementation-example gaps were repaired with nine original examples and answer checkpoints. AB-650 has one actual undated outline despite generic two-version boilerplate; no future version was invented.
AB-210¶
Retain beta/delayed-results and sign-in assessment observations separately. Opportunity custom-field and Close Agent shutdown contradictions remain. Add a competitor-input decision example and an original course-to-lifecycle worksheet; lesson presence does not establish current provisioning.
AB-650¶
Separate Microsoft 365 theme/logo/URL controls from Entra sign-in branding. Distinguish connector-usage inventory from the Power Platform policy plane. The Backup second-year cadence gap remains blocking. Add seven official labs with 450 published minutes and their prepared E7-tenant requirements; public instructions do not supply entitlements or establish November 13 course delivery.
MS-102¶
Update the Connect minimum to 2.6.84.0 and distinguish the Health-agent floor. The April 7, 2027 build-plus-application-authentication requirement is operational planning beyond the November 30 exam retirement. Add bounded Entra PowerShell inventory. Preserve Backup cadence and three-versus-four prerequisite discrepancies; no direct expert-credential successor inferred from replacement training.
PL-400¶
Make the current published four-domain outline explicit while preserving March organization as history. Registration is closed; existing appointments may test through October 30, while new registration is AB-400. Add current code-app command/configuration/CSP/telemetry contracts, Python SDK and background-operation boundaries, standard-harness agent handoff, logical MCP proposal/commit checks, Work IQ surface distinctions, generative-page/agent-flow examples and Monitor alert planning with the scoped hourly aggregation preview. This does not publish a separate AB-400 guide.
AI-500¶
Keep Responses-agent requirements and task-specific A2A setup SDK minimum separate from hosted-agent and protocol versions. Add a protocol/capability failure worksheet and article-to-evidence learning route. Existing memory/guardrail limitations are not silently converted into executed durability, authorization or evaluation tests.
DP-300¶
Accept secure-enclave terminology and add the documented SGX migration boundary. Remove the stale Azure SQL Database compaction preview label while retaining eligible-platform, density/statistics/file-size distinctions. Add proposed one-table SQL permission and GUI inspection examples. SQL authorization and actual Entra sign-in remain independent; no database fixture was executed.
DP-420¶
Promote the current three-domain AI outline and C#/Python audience; mark the old weighted organization as foundation. Add a vector-index feasibility example. Correct the obsolete Agent Kit constructor criticism while retaining its partition-skew caveat. Current course title is Build AI applications with Azure Cosmos DB; older five-domain courses require a gap map.
DP-600¶
Accept the narrow October wording change. Add derived-column/lookup enrichment and the same four-row filter/aggregate exercise through visual, SQL, KQL and DAX routes, with expected East = 20 / West = 30. Add a suspended Direct Lake automatic-update incident and specific two-persona security/paid AI labs. Twelve shared local checks validate arithmetic and SQLite SELECT examples only.
DP-700¶
Accept Eventstream/Eventhouse wording. Replace absolute external-table update-policy prohibitions with the newer narrow asynchronous preview, retaining source-delete, RLS and materialized-view limitations. Add SQL/PySpark/KQL transformations and event-time reasoning, Fabric-specific CDF batch-boundary notes, and monitoring/security labs. Local arithmetic does not establish engine behavior or streaming correction.
DP-750¶
Accept bundle/CLI terminology and separate deletion vectors from clustering. Resolve the former direct-view ABAC conflict with October beta requirements, runtime/workspace prerequisites and metadata-loss risks. Add a schema-boundary CDF example and official lifecycle labs; a bundle deploying placeholder notebooks does not prove a working transformation. Regional automatic-CDF rollout remains a dated plan.
Evidence and validation¶
The initial health pass checked 569 sources: 541 reachable, 26 automation-blocked and two TLS timeouts. Both timed-out Microsoft pages returned HTTP 200 on sequential retry. All 40 new/reused associations checked separately returned HTTP 200. Thirty useful sources were registered; 29 directly evaluated learning entries and 45 selected technical source notes were updated. Reachability, body reading, public outlines, local arithmetic and live execution are separate evidence. Paid lesson interiors, assessments and videos were not consumed.
Two fresh-context read-only reviewers read all ten final guides and complete diffs, checked all 623 current objective bindings and 762 teaching spans, and independently opened selected primary sources. The bounded reviews passed with no new open material findings after two medium repairs: current DP-420 public prerequisites now say C#/Python, and PL-400 includes the code-app-only hourly aggregation preview without promising an hourly delivery SLA. Their complete findings and repair checks are embedded in the operations ledger. This does not complete full rubric-2 semantic audits or human review.
All 199 unit tests, repository validation, reviewed-resource synchronization, strict MkDocs build, generated-site validation and whitespace checks passed. Preservation checks retain all 289 earlier source-review rows, 166 earlier deep-review receipts and 42 AI-audit batches, with only documented objective pointers moved to six byte-identical archives. Eight unchanged objective/status files remain byte-identical. Twelve local Python/SQLite checks validate the new arithmetic examples only. The final report/ledger assembly is followed by another repository and website gate before stage/commit/push. Remote workflows are checked after push; a maintenance-PR Pages build is not a production deployment.
Continue the Microsoft review queue after this group. Remaining source blockers stay visible and are not counted as clean passes merely because bounded research completed.