chore(porch): pir-272 state commits stranded after #305 merged by pseudoseed · Pull Request #309 · pseudoseed/codev

pseudoseed · 2026-08-31T22:13:52Z

The --merged record porch wrote for PR #305 landed on builder/pir-272 after that PR had already merged, so it never reached main. This carries it across.

Contents: codev/projects/272-the-sidebar-tree-needs-every-w/status.yaml — merged: true, merged_at, and the matching updated_at. One file, nothing else.

One thing worth naming

Before merging origin/main in, this branch also carried a regression to another builder's state: it would have deleted merged: true and merged_at from codev/projects/bugfix-298-.../status.yaml and rolled its updated_at backwards from 20:02:35 to 19:55:09.

That was not a hand-edit. This worktree held a copy of that file from before bugfix-298 recorded its own merge, and porch done 272 --merged 305 committed whatever was in the tree alongside its own write. Merging origin/main first resolved it — the file is now identical to main and absent from this diff, which git diff --name-only origin/main..HEAD confirms.

Worth knowing generally: a porch state write commits the tree, not just the file it changed, so a stale sibling status.yaml in a long-lived worktree rides along as a silent revert of somebody else's project state. Related to #233.

Refs #233, #272.

🤖 Generated with Claude Code

The review shipped saying a pairing-issued one-time token is spent on the first sweep tick before anyone spawns, and that an unbounded desktop seed is therefore effectively mandatory. That is wrong. The architect verified against a live server that the token is durable and re-exchangeable — nothing is burned. The real cost is smaller and different: each exchange mints a SESSION, so the sweep accumulates one per server every 30s. A leak, not a spent credential, and it wants a different fix — reuse the access token across ticks rather than provisioning another kind of credential. Tracked in #306. Corrected in place with the wrong version named rather than quietly rewritten, because it was merged and someone may have read it. A merged review is what gets grepped in six months. The underlying mistake is worth stating: I propagated a constraint documented in a comment on ThreadBackendConfig.bootstrapToken as though it were a measurement. It was not, and the live check disagreed with it. Refs #272, #306.

This reverts commit 872d4eb.

pseudoseed · 2026-08-31T22:13:52Z

The --merged record porch wrote for PR #305 landed on builder/pir-272 after that PR had already merged, so it never reached main. This carries it across.

Contents: codev/projects/272-the-sidebar-tree-needs-every-w/status.yaml — merged: true, merged_at, and the matching updated_at. One file, nothing else.

One thing worth naming

Before merging origin/main in, this branch also carried a regression to another builder's state: it would have deleted merged: true and merged_at from codev/projects/bugfix-298-.../status.yaml and rolled its updated_at backwards from 20:02:35 to 19:55:09.

That was not a hand-edit. This worktree held a copy of that file from before bugfix-298 recorded its own merge, and porch done 272 --merged 305 committed whatever was in the tree alongside its own write. Merging origin/main first resolved it — the file is now identical to main and absent from this diff, which git diff --name-only origin/main..HEAD confirms.

Worth knowing generally: a porch state write commits the tree, not just the file it changed, so a stale sibling status.yaml in a long-lived worktree rides along as a silent revert of somebody else's project state. Related to #233.

Refs #233, #272.

🤖 Generated with Claude Code

The review shipped saying a pairing-issued one-time token is spent on the first sweep tick before anyone spawns, and that an unbounded desktop seed is therefore effectively mandatory. That is wrong. The architect verified against a live server that the token is durable and re-exchangeable — nothing is burned. The real cost is smaller and different: each exchange mints a SESSION, so the sweep accumulates one per server every 30s. A leak, not a spent credential, and it wants a different fix — reuse the access token across ticks rather than provisioning another kind of credential. Tracked in #306. Corrected in place with the wrong version named rather than quietly rewritten, because it was merged and someone may have read it. A merged review is what gets grepped in six months. The underlying mistake is worth stating: I propagated a constraint documented in a comment on ThreadBackendConfig.bootstrapToken as though it were a measurement. It was not, and the live check disagreed with it. Refs #272, #306.

This reverts commit 872d4eb.

pseudoseed · 2026-08-31T22:13:52Z

The --merged record porch wrote for PR #305 landed on builder/pir-272 after that PR had already merged, so it never reached main. This carries it across.

Contents: codev/projects/272-the-sidebar-tree-needs-every-w/status.yaml — merged: true, merged_at, and the matching updated_at. One file, nothing else.

One thing worth naming

Before merging origin/main in, this branch also carried a regression to another builder's state: it would have deleted merged: true and merged_at from codev/projects/bugfix-298-.../status.yaml and rolled its updated_at backwards from 20:02:35 to 19:55:09.

That was not a hand-edit. This worktree held a copy of that file from before bugfix-298 recorded its own merge, and porch done 272 --merged 305 committed whatever was in the tree alongside its own write. Merging origin/main first resolved it — the file is now identical to main and absent from this diff, which git diff --name-only origin/main..HEAD confirms.

Worth knowing generally: a porch state write commits the tree, not just the file it changed, so a stale sibling status.yaml in a long-lived worktree rides along as a silent revert of somebody else's project state. Related to #233.

Refs #233, #272.

🤖 Generated with Claude Code

The review shipped saying a pairing-issued one-time token is spent on the first sweep tick before anyone spawns, and that an unbounded desktop seed is therefore effectively mandatory. That is wrong. The architect verified against a live server that the token is durable and re-exchangeable — nothing is burned. The real cost is smaller and different: each exchange mints a SESSION, so the sweep accumulates one per server every 30s. A leak, not a spent credential, and it wants a different fix — reuse the access token across ticks rather than provisioning another kind of credential. Tracked in #306. Corrected in place with the wrong version named rather than quietly rewritten, because it was merged and someone may have read it. A merged review is what gets grepped in six months. The underlying mistake is worth stating: I propagated a constraint documented in a comment on ThreadBackendConfig.bootstrapToken as though it were a measurement. It was not, and the live check disagreed with it. Refs #272, #306.

This reverts commit 872d4eb.

pseudoseed · 2026-08-31T22:13:52Z

The --merged record porch wrote for PR #305 landed on builder/pir-272 after that PR had already merged, so it never reached main. This carries it across.

Contents: codev/projects/272-the-sidebar-tree-needs-every-w/status.yaml — merged: true, merged_at, and the matching updated_at. One file, nothing else.

One thing worth naming

Before merging origin/main in, this branch also carried a regression to another builder's state: it would have deleted merged: true and merged_at from codev/projects/bugfix-298-.../status.yaml and rolled its updated_at backwards from 20:02:35 to 19:55:09.

That was not a hand-edit. This worktree held a copy of that file from before bugfix-298 recorded its own merge, and porch done 272 --merged 305 committed whatever was in the tree alongside its own write. Merging origin/main first resolved it — the file is now identical to main and absent from this diff, which git diff --name-only origin/main..HEAD confirms.

Worth knowing generally: a porch state write commits the tree, not just the file it changed, so a stale sibling status.yaml in a long-lived worktree rides along as a silent revert of somebody else's project state. Related to #233.

Refs #233, #272.

🤖 Generated with Claude Code

The review shipped saying a pairing-issued one-time token is spent on the first sweep tick before anyone spawns, and that an unbounded desktop seed is therefore effectively mandatory. That is wrong. The architect verified against a live server that the token is durable and re-exchangeable — nothing is burned. The real cost is smaller and different: each exchange mints a SESSION, so the sweep accumulates one per server every 30s. A leak, not a spent credential, and it wants a different fix — reuse the access token across ticks rather than provisioning another kind of credential. Tracked in #306. Corrected in place with the wrong version named rather than quietly rewritten, because it was merged and someone may have read it. A merged review is what gets grepped in six months. The underlying mistake is worth stating: I propagated a constraint documented in a comment on ThreadBackendConfig.bootstrapToken as though it were a measurement. It was not, and the live check disagreed with it. Refs #272, #306.

This reverts commit 872d4eb.

pseudoseed · 2026-08-31T22:13:52Z

The --merged record porch wrote for PR #305 landed on builder/pir-272 after that PR had already merged, so it never reached main. This carries it across.

Contents: codev/projects/272-the-sidebar-tree-needs-every-w/status.yaml — merged: true, merged_at, and the matching updated_at. One file, nothing else.

One thing worth naming

Before merging origin/main in, this branch also carried a regression to another builder's state: it would have deleted merged: true and merged_at from codev/projects/bugfix-298-.../status.yaml and rolled its updated_at backwards from 20:02:35 to 19:55:09.

That was not a hand-edit. This worktree held a copy of that file from before bugfix-298 recorded its own merge, and porch done 272 --merged 305 committed whatever was in the tree alongside its own write. Merging origin/main first resolved it — the file is now identical to main and absent from this diff, which git diff --name-only origin/main..HEAD confirms.

Worth knowing generally: a porch state write commits the tree, not just the file it changed, so a stale sibling status.yaml in a long-lived worktree rides along as a silent revert of somebody else's project state. Related to #233.

Refs #233, #272.

🤖 Generated with Claude Code

The review shipped saying a pairing-issued one-time token is spent on the first sweep tick before anyone spawns, and that an unbounded desktop seed is therefore effectively mandatory. That is wrong. The architect verified against a live server that the token is durable and re-exchangeable — nothing is burned. The real cost is smaller and different: each exchange mints a SESSION, so the sweep accumulates one per server every 30s. A leak, not a spent credential, and it wants a different fix — reuse the access token across ticks rather than provisioning another kind of credential. Tracked in #306. Corrected in place with the wrong version named rather than quietly rewritten, because it was merged and someone may have read it. A merged review is what gets grepped in six months. The underlying mistake is worth stating: I propagated a constraint documented in a comment on ThreadBackendConfig.bootstrapToken as though it were a measurement. It was not, and the live check disagreed with it. Refs #272, #306.

This reverts commit 872d4eb.

pseudoseed · 2026-08-31T22:13:52Z

The --merged record porch wrote for PR #305 landed on builder/pir-272 after that PR had already merged, so it never reached main. This carries it across.

Contents: codev/projects/272-the-sidebar-tree-needs-every-w/status.yaml — merged: true, merged_at, and the matching updated_at. One file, nothing else.

One thing worth naming

Before merging origin/main in, this branch also carried a regression to another builder's state: it would have deleted merged: true and merged_at from codev/projects/bugfix-298-.../status.yaml and rolled its updated_at backwards from 20:02:35 to 19:55:09.

That was not a hand-edit. This worktree held a copy of that file from before bugfix-298 recorded its own merge, and porch done 272 --merged 305 committed whatever was in the tree alongside its own write. Merging origin/main first resolved it — the file is now identical to main and absent from this diff, which git diff --name-only origin/main..HEAD confirms.

Worth knowing generally: a porch state write commits the tree, not just the file it changed, so a stale sibling status.yaml in a long-lived worktree rides along as a silent revert of somebody else's project state. Related to #233.

Refs #233, #272.

🤖 Generated with Claude Code

The review shipped saying a pairing-issued one-time token is spent on the first sweep tick before anyone spawns, and that an unbounded desktop seed is therefore effectively mandatory. That is wrong. The architect verified against a live server that the token is durable and re-exchangeable — nothing is burned. The real cost is smaller and different: each exchange mints a SESSION, so the sweep accumulates one per server every 30s. A leak, not a spent credential, and it wants a different fix — reuse the access token across ticks rather than provisioning another kind of credential. Tracked in #306. Corrected in place with the wrong version named rather than quietly rewritten, because it was merged and someone may have read it. A merged review is what gets grepped in six months. The underlying mistake is worth stating: I propagated a constraint documented in a comment on ThreadBackendConfig.bootstrapToken as though it were a measurement. It was not, and the live check disagreed with it. Refs #272, #306.

This reverts commit 872d4eb.

pseudoseed · 2026-08-31T22:13:52Z

The --merged record porch wrote for PR #305 landed on builder/pir-272 after that PR had already merged, so it never reached main. This carries it across.

Contents: codev/projects/272-the-sidebar-tree-needs-every-w/status.yaml — merged: true, merged_at, and the matching updated_at. One file, nothing else.

One thing worth naming

Before merging origin/main in, this branch also carried a regression to another builder's state: it would have deleted merged: true and merged_at from codev/projects/bugfix-298-.../status.yaml and rolled its updated_at backwards from 20:02:35 to 19:55:09.

That was not a hand-edit. This worktree held a copy of that file from before bugfix-298 recorded its own merge, and porch done 272 --merged 305 committed whatever was in the tree alongside its own write. Merging origin/main first resolved it — the file is now identical to main and absent from this diff, which git diff --name-only origin/main..HEAD confirms.

Worth knowing generally: a porch state write commits the tree, not just the file it changed, so a stale sibling status.yaml in a long-lived worktree rides along as a silent revert of somebody else's project state. Related to #233.

Refs #233, #272.

🤖 Generated with Claude Code

The review shipped saying a pairing-issued one-time token is spent on the first sweep tick before anyone spawns, and that an unbounded desktop seed is therefore effectively mandatory. That is wrong. The architect verified against a live server that the token is durable and re-exchangeable — nothing is burned. The real cost is smaller and different: each exchange mints a SESSION, so the sweep accumulates one per server every 30s. A leak, not a spent credential, and it wants a different fix — reuse the access token across ticks rather than provisioning another kind of credential. Tracked in #306. Corrected in place with the wrong version named rather than quietly rewritten, because it was merged and someone may have read it. A merged review is what gets grepped in six months. The underlying mistake is worth stating: I propagated a constraint documented in a comment on ThreadBackendConfig.bootstrapToken as though it were a measurement. It was not, and the live check disagreed with it. Refs #272, #306.

This reverts commit 872d4eb.

pseudoseed · 2026-08-31T22:13:52Z

The --merged record porch wrote for PR #305 landed on builder/pir-272 after that PR had already merged, so it never reached main. This carries it across.

Contents: codev/projects/272-the-sidebar-tree-needs-every-w/status.yaml — merged: true, merged_at, and the matching updated_at. One file, nothing else.

One thing worth naming

Before merging origin/main in, this branch also carried a regression to another builder's state: it would have deleted merged: true and merged_at from codev/projects/bugfix-298-.../status.yaml and rolled its updated_at backwards from 20:02:35 to 19:55:09.

That was not a hand-edit. This worktree held a copy of that file from before bugfix-298 recorded its own merge, and porch done 272 --merged 305 committed whatever was in the tree alongside its own write. Merging origin/main first resolved it — the file is now identical to main and absent from this diff, which git diff --name-only origin/main..HEAD confirms.

Worth knowing generally: a porch state write commits the tree, not just the file it changed, so a stale sibling status.yaml in a long-lived worktree rides along as a silent revert of somebody else's project state. Related to #233.

Refs #233, #272.

🤖 Generated with Claude Code

The review shipped saying a pairing-issued one-time token is spent on the first sweep tick before anyone spawns, and that an unbounded desktop seed is therefore effectively mandatory. That is wrong. The architect verified against a live server that the token is durable and re-exchangeable — nothing is burned. The real cost is smaller and different: each exchange mints a SESSION, so the sweep accumulates one per server every 30s. A leak, not a spent credential, and it wants a different fix — reuse the access token across ticks rather than provisioning another kind of credential. Tracked in #306. Corrected in place with the wrong version named rather than quietly rewritten, because it was merged and someone may have read it. A merged review is what gets grepped in six months. The underlying mistake is worth stating: I propagated a constraint documented in a comment on ThreadBackendConfig.bootstrapToken as though it were a measurement. It was not, and the live check disagreed with it. Refs #272, #306.

This reverts commit 872d4eb.

pseudoseed added 2 commits August 31, 2026 16:12

chore(porch): 272 PR#305merged
d61189d

Merge remote-tracking branch 'origin/main' into builder/pir-272
34f15c2

pseudoseed mentioned this pull request Aug 31, 2026
A porch state write commits the whole tree, so a stale sibling status.yaml silently reverts another project's state #310
Open

pseudoand others added 2 commits August 31, 2026 16:14

Revert "docs(review): correct the token claim in the merged#272review"
8d521fe
This reverts commit 872d4eb.

pseudoseed merged commit 0e05bf2 into mainAug 31, 2026
9 checks passed

pseudoseed added 2 commits August 31, 2026 16:12

chore(porch): 272 PR#305merged
d61189d

Merge remote-tracking branch 'origin/main' into builder/pir-272
34f15c2

pseudoseed mentioned this pull request Aug 31, 2026
A porch state write commits the whole tree, so a stale sibling status.yaml silently reverts another project's state #310
Open

pseudoand others added 2 commits August 31, 2026 16:14

Revert "docs(review): correct the token claim in the merged#272review"
8d521fe
This reverts commit 872d4eb.

pseudoseed merged commit 0e05bf2 into mainAug 31, 2026
9 checks passed

pseudoseed added 2 commits August 31, 2026 16:12

chore(porch): 272 PR#305merged
d61189d

Merge remote-tracking branch 'origin/main' into builder/pir-272
34f15c2

pseudoseed mentioned this pull request Aug 31, 2026
A porch state write commits the whole tree, so a stale sibling status.yaml silently reverts another project's state #310
Open

pseudoand others added 2 commits August 31, 2026 16:14

Revert "docs(review): correct the token claim in the merged#272review"
8d521fe
This reverts commit 872d4eb.

pseudoseed merged commit 0e05bf2 into mainAug 31, 2026
9 checks passed

pseudoseed added 2 commits August 31, 2026 16:12

chore(porch): 272 PR#305merged
d61189d

Merge remote-tracking branch 'origin/main' into builder/pir-272
34f15c2

pseudoseed mentioned this pull request Aug 31, 2026
A porch state write commits the whole tree, so a stale sibling status.yaml silently reverts another project's state #310
Open

pseudoand others added 2 commits August 31, 2026 16:14

Revert "docs(review): correct the token claim in the merged#272review"
8d521fe
This reverts commit 872d4eb.

pseudoseed merged commit 0e05bf2 into mainAug 31, 2026
9 checks passed

pseudoseed added 2 commits August 31, 2026 16:12

chore(porch): 272 PR#305merged
d61189d

Merge remote-tracking branch 'origin/main' into builder/pir-272
34f15c2

pseudoseed mentioned this pull request Aug 31, 2026
A porch state write commits the whole tree, so a stale sibling status.yaml silently reverts another project's state #310
Open

pseudoand others added 2 commits August 31, 2026 16:14

Revert "docs(review): correct the token claim in the merged#272review"
8d521fe
This reverts commit 872d4eb.

pseudoseed merged commit 0e05bf2 into mainAug 31, 2026
9 checks passed

pseudoseed added 2 commits August 31, 2026 16:12

chore(porch): 272 PR#305merged
d61189d

Merge remote-tracking branch 'origin/main' into builder/pir-272
34f15c2

pseudoseed mentioned this pull request Aug 31, 2026
A porch state write commits the whole tree, so a stale sibling status.yaml silently reverts another project's state #310
Open

pseudoand others added 2 commits August 31, 2026 16:14

Revert "docs(review): correct the token claim in the merged#272review"
8d521fe
This reverts commit 872d4eb.

pseudoseed merged commit 0e05bf2 into mainAug 31, 2026
9 checks passed

pseudoseed added 2 commits August 31, 2026 16:12

chore(porch): 272 PR#305merged
d61189d

Merge remote-tracking branch 'origin/main' into builder/pir-272
34f15c2

pseudoseed mentioned this pull request Aug 31, 2026
A porch state write commits the whole tree, so a stale sibling status.yaml silently reverts another project's state #310
Open

pseudoand others added 2 commits August 31, 2026 16:14

Revert "docs(review): correct the token claim in the merged#272review"
8d521fe
This reverts commit 872d4eb.

pseudoseed merged commit 0e05bf2 into mainAug 31, 2026
9 checks passed

pseudoseed added 2 commits August 31, 2026 16:12

chore(porch): 272 PR#305merged
d61189d

Merge remote-tracking branch 'origin/main' into builder/pir-272
34f15c2

pseudoseed mentioned this pull request Aug 31, 2026
A porch state write commits the whole tree, so a stale sibling status.yaml silently reverts another project's state #310
Open

pseudoand others added 2 commits August 31, 2026 16:14

Revert "docs(review): correct the token claim in the merged#272review"
8d521fe
This reverts commit 872d4eb.

pseudoseed merged commit 0e05bf2 into mainAug 31, 2026
9 checks passed

Conversation

pseudoseed commented Aug 31, 2026

One thing worth naming

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

pseudoseed commented Aug 31, 2026

One thing worth naming

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

pseudoseed commented Aug 31, 2026

One thing worth naming

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

pseudoseed commented Aug 31, 2026

One thing worth naming

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

pseudoseed commented Aug 31, 2026

One thing worth naming

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

pseudoseed commented Aug 31, 2026

One thing worth naming

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

pseudoseed commented Aug 31, 2026