Hive vs Reclaim.ai

Hive publishes its changelog as same-day clusters of single-feature entries — one on August 19, seven on August 18, three on August 14. The newest lets reviewers choose which comment types an AI proofing pass generates and attach custom instructions to it. Behind it sits the August 18 batch, split between planning integrity (time estimates surviving assignee changes, a rebuilt Unsubmitted Timesheets view, reorderable Gantt columns), governance (a default restricted-member role across every onboarding path, custom-field edits in activity feeds), and small chat conveniences.

Three threads run in parallel and all point at larger deployments. Planned time is being made trustworthy, so hours stop vanishing on reassignment. Administrative control is being made to scale, with least-privilege defaults that hold across SAML, SCIM, invite links and domain auto-join. And AI proofing keeps gaining operator controls rather than new intelligence — after learning from reviewer feedback and auditing prior rounds, it now takes scope filters and instructions. The pattern is an AI feature being made governable by the teams that have to sign off on its output.

The proofing controls and the audit trail both look mid-build: per-reviewer instruction presets and activity coverage for more object types are the obvious next steps, at the same twice-monthly batch cadence.