mixpeek/amux is an open-source control layer that runs AI coding agents (Claude Code, Codex, Gemini) in parallel. It divides work on a shared board and manages schedules, recurring runs, model swapping, and automatic recovery, and it ships as a single Rust binary under the MIT license. The November 2, 2026 commit 06483f6 added only 11 lines to frustrations.md and did not touch any code.
The added entry is a boundary error that occurs when moving a card into the review state. In amux, workers are grouped, and messages between workers can only travel within the same group. Yet the `amux board review AUTHO-1 --checked ...` command printed "the card moved to review, but the REVIEWER WAS NOT TOLD ... worker group boundary refused" and then ended successfully. The card was left in the review state with only the reviewer assigned, and the request reached nobody; afterward, neither the author nor the reviewer was notified. When the author tried to post the issue to a separate board, amux-frustrations, it was rejected with cross_board_delegation_forbidden, so the AUTHO-1 card remaining on their own board is exactly this entry.
That said, this commit is a report, not a fix. The entry lists STATUS: open and SEVERITY: blocks, and the three items in its FIX section (fail the transition when a reviewer outside the group is assigned, mark an unnotified review as a reviewer-waiting state rather than a normal review, and give CC_RECEIVE_ANY=1 to the designated cross-lane reviewer) are also only at the proposal stage. The commit message states that the original was 0603fe490, an ad hoc branch push by a demo worker, and that it can be reverted with git revert. So what remains is the reproduction procedure and the error message. That message is stored in a placeholder state where names are not filled in, like author [{}] -> reviewer [{}], so the actual session cannot be known. How often this problem occurs, and whether the same group boundary rule keeps blocking other commands as well, cannot be judged from this single entry alone.