Release notes
Public updates
Tangle provides infrastructure for running and evaluating AI agents. For teams evaluating or operating those services, these notes list public changes from newest to oldest so you can check what changed before relying on a service, policy, benchmark, or guide. For code-level changelogs, inspect the Tangle GitHub organization.
Benchmarks page returns without rankings
The benchmarks page is public again. It lists every suite Tangle measures, what each suite grades, and how much of its last run was actually graded. No ranking is published: the three publishing gates are computed from each run record when the site is built, and every current run fails or cannot prove at least one of them.
Benchmark boards removed
Removed the agent benchmark boards. Some runs had no recorded result or failed before the agent made a model call, and no board had enough tasks and repeated runs to give a stable ranking. A board will be published again only when every run completes and receives a valid grade.
Service transparency update
Added pages where customers can check current service health, review security practices, and inspect current privacy and service terms.
Service statusSecurity informationPrivacy PolicyTerms of Service
Website and benchmarks update
Refreshed the agent benchmark boards and published the method for comparing whether coding agents implement current public API behavior.