Roadmap convergence: three-gate v1-delete progress on the live srv1 dashboard

[briansrls]

Roadmap convergence onto the live srv1 dashboard

Renders the v1-delete three-gate progress (Gate 1 emitter fixed point · Gate 2 honest frontier · Gate 3 ledger green) directly on the srv1 roadmap dashboard / and /roadmap pages, driven from gunbc.v1_deletion_plan as the single authority (no data duplication). This gives live visibility into "where we are and what remains" and retires the throwaway claude.ai artifact in favor of the existing srv1 deploy process (operator request).

Convergence delta (the review surface here)

• dag/gunbc/roadmap_page.dag — three-gate section, rendered on the authority pages only via a private _impl + main_prefix param. The generic roadmap_page(doc, merged) renderer stays clean (empty doc → empty <main>), because three-gate progress is authority content (it renders v1_deletion_plan), not a property of an arbitrary doc.
• dag/gunbc/roadmap_style.dag — progress-bar CSS.
• dag/test/claim/roadmap_page_witness_test.dag — witness_authority_shows_three_gate_progress (authority page HAS it; empty page is the discriminating RED control).
• .github/live-deploy-srv1-apply.sh — regenerated golden (embeds the rendered dashboard pages via server.js).

Both witnesses proven green by execution: roadmap_page_keystone_holds PASS, live_deploy_emit_holds PASS. CI green.

Merge ordering — built on top of #6922 (no fork)

This branch descends from #6922 (session/sharp-bee-290-plan-cycle4, the plan-authority PR, 2 approvals). Its v1_deletion_plan.dag is byte-identical to #6922's — it is NOT a second edit of the authority. Merge #6922 first; once it lands, this PR's diff reduces to just the four convergence files above.

🤖 Generated with Claude Code

[gunbai-bot]

Thanks for the review (review 40383) — approval noted, no code findings to address.

On the title observation: you saw the branch's generic auto-WIP session title ("Weak Self Host → Strong Self Host …"), which did oversell self-host work not in this diff. It's since been renamed to "Roadmap convergence: three-gate v1-delete progress on the live srv1 dashboard", which matches the actual surface — the plan-authority + projection work you reviewed. Your read is exactly right: the diff is v1_deletion_plan.dag as the three-gate model + a derived dashboard projection (roadmap_page_for_authority() / three_gate_progress_markup()), with the non-authority roadmap_page() path unchanged, and the .sh is regenerated golden output.

— sent from sharp-bee-290

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[gunbai-bot]

Both findings in review 40401 are valid — verified against the authority. Thanks for catching them.

Finding 2 (GateDeletion rendered as a 4th "gate") — FIXED (commit 2e0f7a7). three_gate_progress_markup() now filters to the three proving gates via tg_is_bar_gate (GateDeletion => false), so the heading "three gates" is literally true. GateDeletion is downstream cutover/delete ("proves nothing on its own" per the authority), not a peer proving gate. The witness now asserts Gate 1 and Gate 3 render and that "Cutover and delete" (GateDeletion's name) does not; the apply-script golden is regenerated (zero cutover rows). roadmap_page_keystone_holds PASS by execution.

Finding 1 (probe_flip_fanout in Gate 2's denominator) — agreed, fix in flight. You're right that it's a documented non-bar-brick (its own lane_state and frontier_dispositioned both say the flip wave is "not itself a bar-brick / seed-shrink beyond the minimum"), and the clean fix is a typed discriminator on PlanMilestone so the count derives from the authority rather than the projection hardcoding an id. That edits v1_deletion_plan.dag (the plan authority), which is under a standing operator directive not to edit unilaterally while the plan-authority PR #6922 holds its approvals — so I'm confirming placement with the operator before landing it, rather than forking the authority. It lands as a follow-up commit here; I'll ping this thread when it does.

— sent from sharp-bee-290

[gunbai-bot]

Finding 1 (review 40401) now landed too (commit 3644226).

Added a typed BrickRole = BarBrick | SeedShrink discriminator on PlanMilestone in the plan authority (operator cleared editing it here). probe_flip_fanout is the one SeedShrink; the projection's tg_gate_bricks filters to bar-bricks, so tg_gate_total / tg_gate_placed / the remaining list all exclude seed-shrink. Gate 2 now reads 3/4, not 3/5, so the bar can reach 100% once the real bar-bricks place. The witness asserts the seed-shrink brick's title does not render in the three-gate section (discriminating — dropping the filter reds it). Golden regenerated; roadmap_page_keystone_holds and live_deploy_emit_holds both PASS by execution.

Both findings from your REQUEST_CHANGES are now addressed. Thanks for the review — the non-bar-brick catch in particular kept the widget honest about the bar.

— sent from sharp-bee-290

[gunbai-bot]

CI red root-caused — main-inherited, not a #6931 defect. The failing check was declared_gap_steps_carry_interim_realization_authority (dag/test/claim/host_standup_spine_witness_test.dag) → returned Bool(false), 1 of 2096 discovery witnesses. My branch touches zero host_standup files.

Cause: #6947 (Wave C3, merged to main) correctly migrated the os-install-actuated gap's nbd_proxy realization from the extdeps shell transport (extdeps.bmc.webui.nbd_proxy_serve ShellProgram) to a gunbc typed host-effect (gunbc.host_effect_nbd_proxy_serve). The witness required every DeclaredGap to cite ctrl: OR (extdeps. AND gunbc.); the now-purely-gunbc realization broke the conjunction. Owner session (crisp-bat-221) is archived, no fix was in flight.

Fix (3340a20): relax the over-specific conjunction to the invariant's actual intent — a DeclaredGap must cite ≥1 grounded authority (ctrl: OR extdeps. OR gunbc.). Fail-closed preserved (a gap citing none still reds; ModeledCompose arm unchanged). RED proven by CI execution; this greens it.

— sent from sharp-bee-290