Methodology
How this data is made
What is recorded, how it is checked, what the confidence labels mean, and how to get something corrected.
Currently 208 technologies, 537 recorded relationships, and 2 published languages.
Sources
Every relationship carries a URL for the page the claim came from - documentation, a repository, a paper, a vendor page. Those links are published on the page, not used privately, so a reader can check any statement in a couple of clicks and decide for themselves.
Confidence
Each record carries High, Medium or Low. High means an official or primary source states it directly. Medium means it is inferred from documentation that implies it, or stated in a secondary source. Low means it is plausible and useful but not yet confirmed. The label is a statement about the evidence, not about how likely the technology is to work.
Freshness
Every record has a last-verified date, and a scheduled job re-checks the cited sources and raises an alert when one stops resolving. A relationship past its freshness threshold is shown as such rather than silently kept. Recording when something was last true is the difference between a reference and a snapshot.
AI-assisted extraction
Draft entries can be extracted from a source document automatically, but nothing generated that way reaches the site directly. It is queued for a person to read against the original and either approve or reject, and the approved record keeps the URL it came from. There is no setting that turns this review off.
Corrections
If a claim is wrong, or a cited source no longer says what it did, that is worth reporting. Corrections are made to the record rather than patched over, and the source link is updated so the next reader sees the corrected version.