comparison · c_25172cf8caa0 (tommy-mor) vs c_e4c160a5feb7 (tommy-mor)
Side A fixes a real concurrency bug (shared browser sentinel delegate blocking multi-user votes) with a coherent design change across server, CLI, types, and tests, including a new integration test proving the fix. Side B is a minor cosmetic HTML cleanup removing headers/labels with no functional change or test coverage, offering far less lasting value.
A fixes a real multi-user correctness bug by removing shared browser sentinel delegates (PendingSession.agent becomes Option, votes/posts use no delegate, authorship simplified) with integration coverage; B only deletes cosmetic forum chrome (compose heading/help text and thread subtitle/garden link).
Side A fixes a substantive correctness issue by removing shared browser sentinel delegates, changing pending sessions to use an optional delegate for browser logins, updating attribution logic, and adding integration tests that verify multiple human users can vote without erroneous shared AgentBound conflicts. Side B only makes minor UI cleanup by removing or relocating headings and help text in the compose view, with no evident behavioral improvement.
comparison · c_e4c160a5feb7 (tommy-mor) vs c_6f04dcb2e38c (tommy-mor)
Side B removes a significant chunk of legacy dual-path projection logic (GitDiscovery-based commit/epoch rendering), simplifying the codebase to a single Evidence-envelope model, with corresponding test updates and required-field cleanup in the Emission dataclass — a substantive architectural simplification. Side A is a small, purely cosmetic HTML/view tweak (removing a heading and reordering a div) with no functional or structural impact.
B removes substantial legacy projection paths (_legacy_commit_row, GitDiscovery fallbacks on epoch/commit pages) and enforces Evidence-envelope-only rendering with matching test updates, a lasting schema/architecture cleanup. A only rearranges a few forum HTML nodes (heading/muted copy and error div placement) with no functional or design impact.
Side B makes a substantive architectural change by removing legacy GitDiscovery projection paths and requiring Evidence-envelope-backed data throughout commit and epoch pages, simplifying the data model and enforcing the current schema. It updates required Emission fields, rewrites page logic to consume evidence events directly, removes legacy compatibility code and tests, and adjusts integration tests accordingly, whereas Side A only removes or repositions UI elements in the forum view with little lasting functional impact.