How a page reaches the library
Each article addresses a distinct task, records sources and evidence, and passes a technical and editorial check before its published record becomes a page. We use AI assistance, including Codex, to draft, inspect and test content. The initial technical review is AI-assisted; a review date does not imply a named independent human certification.
A change to an article invalidates its previous approval. Titles, canonical URLs, dates, source links, related pages, product facts and any attached examples are checked before release. Unfinished drafts are excluded from public routes, the sitemap and our optional text index.
Product facts and source documents
Pricing, duration and allowance come from the service configuration. Protocol explanations link to standards or the relevant client’s documentation. Each article shows its own substantive update and review dates. Checking a reference without changing the article does not pretend the whole page is newly published.
We do not publish keyword variants that claim unsupported locations, retailer acceptance, purchasing-bot compatibility, IP reputation or guaranteed purchases. A future compatibility statement needs a dated, reproducible test and an explicit scope.
What the setup fixtures prove
Our initial client tests use an isolated HTTP proxy, an HTTP origin and an HTTPS origin on the same machine. A locally trusted test certificate preserves TLS verification. Synthetic credentials contain characters that exercise encoding; no customer credentials are used. The origin is checked for leaked proxy authentication headers.
The fixture checks authenticated forwarding and CONNECT, wrong-credential rejection, a denied CONNECT, an origin denial after a successful tunnel, and a bounded stalled-connection timeout. Versions and sanitized results are available in the test evidence file.
These tests do not prove a live cloud public exit IP, a retailer or bot integration, production availability, latency, throughput, purchase success, or a service-level guarantee. The setup pages make that boundary explicit. We will publish live benchmarks only with a real dataset, sample size, method and limitations.
Corrections and review requests
If a command fails on a listed version or a product statement appears wrong, open a support ticket or email hello@flintblocks.io. Include the page URL, client version, timestamp and a redacted description. Do not include proxy passwords, tokens, cookies or personal account data.
We correct material errors and update the affected article’s date. A stale or unsupported claim can be removed while its evidence is rechecked. Published examples are informational connection diagnostics; the service terms define the rental.