Salesforce Platform App Builder deep review — September 29, 2026

The guide maps 26 saved objective statements, answers 40 original prompts and adds 28 executed local checks. The current browser-rendered outline has the same substantive objectives and weights, with one editorial difference. An official maintenance inconsistency and several source-access limits remain recorded.

Canonical comparison and maintenance

The canonical Help article returned a loading/CSS-error shell to the automated fetch. The monitor correctly reports manual review. Its complete browser-rendered main text was reviewed; the five sections contain 5/4/8/5/4 objectives at 18/20/32/17/13 weights. The only objective-text difference is an added “the” before AgentExchange in the second bullet. Existing objective and lifecycle baseline files were retained; the evidence records the manual comparison and does not claim an automated current-source hash match.

The exam article's three-times-yearly maintenance sentence conflicts with the dedicated maintenance policy and schedule. Their English bodies were readable through German-locale wrappers and specify an annual Winter badge for App Builder. The schedule lists December 11, 2025 availability and December 4, 2026 due date for Winter ’26, in Pacific time, subject to change. The guide directs holders to the requirement assigned in their profile while retaining the conflicting source wording. No credential account or support request was used.

The August 27 Salesforce Admins interview confirms deeper Flow creation, monitoring and troubleshooting and foundational Agentforce. It does not establish the previously cited exact August 21 launch date. The formerly supporting independent article currently exposes only a generic title, so that exact date is no longer asserted as freshly verified.

Technical changes and actual local execution

Derived-field audiences receive explicit treatment: a visible cross-object formula can expose a parent value without parent-record access, and roll-up outputs can reveal aggregates of hidden detail fields. Proposed activities test the resulting field audience and relationship deletion behavior. Approval guidance distinguishes coordinated stage/process locking from parallel steps that each attempt to lock the same record. Administrator/allowed approver edits also mean locking is not an immutable review snapshot.

Deployment guidance distinguishes validation from actual deployment and inspects exact component/status results, including nonproduction partial success. A later deployment invalidates prior quick-deploy eligibility. API 68 context enforcement is versioned and does not substitute the chatting customer's identity for the actual running user.

The original Python workbook executes standard-library SQLite foreign keys, CASCADE/SETNULL deletion, joins, explicit commit/rollback and a declared dependency graph using graphlib. Three details totaling 250 cents become a false 500 after joining two notes; SUM(DISTINCT) instead produces 150 by discarding an equal-value detail. Aggregating by request preserves 250. Invalid child/parent-budget changes roll back, and three failure paths distinguish a handled commit, current-transaction rollback and an earlier committed update. Graph checks cover prerequisite order, supplied available dependencies, missing dependencies and cycles.

All 28 checks executed locally. These are bounded design demonstrations, not a Salesforce emulator or runtime test. Native roll-up/currency/sharing/RecycleBin behavior, Flow transactions, approval locks, actual target dependency discovery, concurrency, crash recovery and production performance remain untested. In-memory connections close after each fixture; no new dependency, account, network call or service change was needed.

Learning catalog and reading limits

The current preparation trail lists 39 hr 20 min: 20 minutes of overview plus 9h25/4h50/10h45/11h20/2h40 across the five topic groups. Card metadata includes AgentExchange and Agentforce admin setup; badge interiors were not all reviewed. Winter ’26 maintenance lists 30 minutes, split 5-minute introductory unit and 25-minute activity. The main intro was read, but no assessment answers or activity completion were used. Its linked Create Records release page was unreadable; precise API 65 output-type wording was not adopted.

The O’Reilly bootcamp public outline names Michaela Grofčík and eight timed sections totaling 300 minutes, including 10-minute wrap-up and excluding untimed breaks. No current event date or lesson/practice content was verified. Udemy and the O’Reilly video endpoint returned 403. YouTube exposed a “5 Minute” title and footer only; actual runtime, date and video were not verified. Focus on Force remains a general catalog reference, with no current App Builder product-quality claim.

All 24 direct sources have HTTP receipts: 22 HTTP successes and two blocked. Several successful responses contain only loading shells or titles. Browser fallbacks and selected/complete reading boundaries are recorded separately; successful HTTP transport is not counted as readable content. The three-column learning table separates current public durations, earlier unverified observations and our planning estimates.

Evidence and remaining work

The operational ledger is ADLC_Docs/operations/2026-09-29-salesforce-platform-app-builder-deep-review.json. It retains prior records, original hashes, manual comparison, source receipts, reading limits, objective mapping, local execution and validation results. Eight org activities, independent human review, maintenance-source clarification and inaccessible commercial metadata remain pending. The five reserved GitHub guides and disabled notification pilot remain untouched.