Cloud Digital Leader deep review — September 29, 2026¶
Same-context AI review; independent human review pending. Study guide.
Scope and lifecycle¶
Read the actual exam PDF, seven pages, and the study PDF, fifteen pages. The August 12, 2026 outline contains 69 detailed considerations under 14 numbered objectives: groups 8/14/12/10/15/10, weights 18/18/18/18/18/10. Both objective and lifecycle monitor hashes remain unchanged; accepted snapshots were not rewritten. PDF byte counts and hashes are retained in the operational receipt.
The certification page still distinguishes a standard exam and shorter renewal exam with three-year validity from designated Google Skills renewal with one-year validity. Eligibility and account-linking requirements are now explicit. No later blueprint replacement was announced on the checked page.
Corrections and original teaching¶
The April 22 announcement, by Brian Delahunty and Michael Gerstenhaber, was fully read. It distinguishes the developer Agent Platform, evolving Vertex AI, from the employee Gemini Enterprise app. The guide now separates those roles and explains that agent workflow construction does not necessarily train model weights. Marketing claims are not treated as measured enforcement or universal availability.
Current Cloud Run documentation broadens the old request-only description: services, finite jobs and other execution forms have different behavior. Worker scaling and disposable local state are qualified; additional product forms are supporting context, not new exam objectives. Other corrections cover online Archive access, Autoclass as management rather than a class, billing duration versus retention, default strong Firestore reads, cumulative IAM allow grants, and external Cloud SQL clients versus service location.
Model Armor detection, an LLM refusal and an enforced integration block are distinct. The guide requires correlated evidence and checks modality coverage. AI Protection visibility remains subject to activation scope and tier.
Spend cap budgets are a separate preview capability from alerts-only budgets. Their project/service scope, gross estimated cost basis, enforcement delay, continuing charges and manual lifting prevent an exact whole-account spending guarantee. No budget, cap or notification was configured. DWS pricing distinguishes reserved-duration Calendar mode from usage-based Flex-start billing and flags November 2026 changes; no rate is embedded.
An exact public standard-library Python worksheet passed 26 local checks. It demonstrates a weighted SLI versus misleading average percentages, explicit successful-and-fast request populations, lower total cost with higher cost per completed order, simple assumed forecasts and delayed/scoped cost signals. Invalid or empty evidence is rejected. The public code is hash-matched; these calculations do not emulate Google billing, enforcement or SLO services.
The guide has 44 answered checks, three integrated scenarios and stronger acceptance/negative cases for eight proposed decision labs. No cloud resource, API, model, identity policy, billing control or recovery operation was tested live.
Learning resources¶
Google Skills exposes six activities and a relative update eight days earlier; its current course durations were not exposed. The Google-authored Coursera series has six course rows totaling ten hours, while its FAQ and landing page give different broader estimates. Its public outline now explicitly mentions agentic AI, AI Hypercomputer and Model Armor. Whizlabs exposes seven modules and 7.5–8 video hours versus about twenty total study hours; public titles retain Cloud Functions and do not establish every August addition. These are public metadata comparisons, not reviews of paid interiors or assessments.
O'Reilly and Udemy access was blocked, so inherited durations/update dates are qualified as unverified. YouTube and sample-form fetches did not expose full content. Planning estimates are labeled and Places to learn remains the final section.
Validation and limits¶
The operational receipt records local worksheet checks, catalog consistency, repository tests, repository validation, strict site generation, generated-site checks and diff checks after they run. Source freshness is current. The program outcome remains reviewed with blockers because proposed cloud verification and independent human review remain outstanding. Two initially missing documentation paths were corrected and their failed attempts are preserved separately.