An opaque control separates ordinary alpha conversion from a deliberate one-byte canvas readback alteration.
When does canvas byte inequality establish readback alteration, and what control excludes ordinary alpha conversion?
Ten requests through the reference Worker show where clearance is checked, which requests reach the origin, and why an image header does not admit an export.
Eight controlled rustls handshakes show why a hybrid offer alone does not predict failure. The mixed ClientHello carries a separate classic key share.
A two-CPU quota left Chromium reporting fourteen processors. Restricting affinity changed the count to two. Nine sessions show why throughput needs its own measurement.
An IP-in-IP packet experiment compares forwarding with TCP termination on the same delayed link.
A cost and protocol comparison of SMS verification and network-informed mobile escalation, with executable pricing and measurement-commitment examples.
A packet experiment separates a terminating proxy’s TCP handshake from the browser’s application response, with a direct-client counterexample.
The decisive code, observations or reasoning are in the article body.
Claims name their premises and the cases that break them.
Test conditions and results carry enough detail to inspect.
A separate technical review checks the evidence before publication.
The public demo returns the verdict and its named findings.