How we test
Every number on this site comes from the same place: an automated harness that resolves real TeraBox share links and records what happened. No figure is written by hand, and no figure is written by a language model. This page explains the method so you can judge the numbers rather than trust them.
What the harness measures
- Time to file list
- From submitting a share link to the file list being readable. Reported as a median across repeated runs, never as a single best case.
- Download time and throughput
- Measured on a sample file of a stated size, over a stated connection. Both the size and the connection appear in the table so the figure means something.
- Links that opened
- The share of test links that produced a usable file list, across all eight supported domains.
- Error distribution
- Which failure causes actually occur, taken from the error catalogue our resolver returns rather than from guesswork.
How figures are reported
- Every figure carries its method: a median states how many runs, a rate states how many links.
- A single observation is labelled as one observation, never presented as a median.
- Failures are published. A domain that opened 92% of links is reported at 92%, not rounded up or left out.
- Figures are never converted into a different form. A measured 95.7% is not restated as "about 96%" or "roughly 1 in 20".
What the harness cannot measure
Some behaviour belongs to the operating system rather than to any link. Where Safari saves a download, whether the Photos app accepts an MKV file, whether a browser can resume an interrupted transfer — none of that can be established by resolving links. Those claims come from a separate fact base, each entry either supported by primary documentation from Apple, Google, or the IETF, or verified by hand and marked as such.
Rollout status. The harness is being brought online across the site, and figures are replaced with live measurement runs as each page is covered. Pages that carry sample figures are progressively swapped to measured ones without their text being rewritten, because the text quotes the figures rather than restating them.
How the writing works
Article text is drafted with AI assistance, working only from the measured figures and the verified fact base. The drafting system cannot see anything else, and an automated gate rejects any sentence containing a number that has no counterpart in the underlying data. Our editorial policy covers the rest.
Corrections
If a figure looks wrong, tell us and we will re-run the measurement. Corrections are made in the open and recorded in the update history at the bottom of each page. Reach us through the contact page.