The argument, with the mechanism
Don't trust, verify.
A source link gives you something to check. Practice questions carry documentation references, while source checks help identify changes that need review. This does not guarantee that every answer is current.
One. Source records make changes inspectable
The catalog records source URLs and fingerprints for documentation pages and supporting passages. These records support automated checks and review; their presence alone does not establish that a passage proves an answer.
Source checks flag changed sources for review and can propose quarantine or corrections. Reviewed catalog updates are published separately: detection does not automatically withdraw a question from the live site.
Two. Every question knows which objective it tests
The exam guide is not decoration here; it is the index. Each question records the exam guide version it was written against, the section it belongs to and the specific objective it tests, and none of those may be blank. That is why a report card can tell you which objectives cost you marks instead of handing you one number and a shrug.
It also means you can see the coverage for yourself. A blueprint area with no questions shows up as a gap on your own screen.
Three. It matches the exam's shape without reproducing its content
Machine-written questions have a habit: the correct answer ends up longer and more careful than the wrong ones, and a candidate who spots that stops reading the scenario and starts counting characters. Practice like that teaches a tell, then the real exam takes it away.
The checked-in Professional Cloud Architect reference uses the vendor's published sample questions, 137 of them. In that cohort the correct option is the longest just over half the time. We use those measurements to check generated questions for answer-length clues. Other exams need their own reference; sample coverage and quality checks can have gaps.
What is copied is the statistics. Not a stem, not an option, not a scenario. The exam's published samples tell us what a question of this exam looks like; the documentation tells us what is true; the questions are ours.
And three things it refuses to do
The boundaries we aim to keep explicit
No forecast of your result
No likelihood, no forecast score, no verdict that you are prepared. A test greps these pages for that family of phrasing and fails the build when it appears, because the tempting place to write it is exactly the page a learner most wants to believe.
What you get instead is evidence: what you answered, unaided, and how much of the blueprint it covered.
No dumps, and no locked content
Policy excludes paid, leaked and access-controlled exam material. Community sources are only ever read for signal, which topics come up and how often, never for wording.
And the free tier is the product, not a sample of it. No advertising is currently enabled. Planned support options do not unlock questions. The prices are published for future support. Payments are not switched on yet.