Content pipeline handoff — 2026-09-23

Outcome and scope

Establish a small, workable publishing process using the 2RCon backlog and upcoming Emergent Dialog interviews as cases, then test a bounded improvement. Content strategy is parked after its revised SCQ and nine-question tree. Do not reopen the 6–12 month programme as a prerequisite for operational progress.

Use Beads for task specifications, ownership mode, estimates, dependencies, and live state. This note provides starting context and execution boundaries; it is not a second task queue. Parent tracker: lip-93f.2. Handoff task label: handoff:content-pipeline.

Start here

Read the repository AGENTS.md and roadmap, then:

Pipeline project and compact issue tree.
Initial case evidence.
The current Bead description before each task. The revised Bead specifications supersede the earlier generic sequence that required full strategy input and a separate framing review before proposing a workflow.

GitHub project #38 is accessible through gh following the permission refresh. The last read returned 37 entries. Existing 2RCon records are #215: release approach, #160: post-production, and #192: marketing placeholder. Historical checklists show mixed edit/upload states; verify selected records rather than treating their dates or checkmarks as current reality. Drive/catalogue links are in the evidence note.

The exact Emergent Dialog interview records still need identification. Do not assume they are the four Debold episodes in comms #190 or confuse them with workshop/residency promotion. Missing access or identity should be recorded as a specific question; it should not prevent useful 2RCon work.

Execution modes and model recommendation

who:ai: a worker can complete the specified research, synthesis, drafting, or evaluation independently with available evidence. Default recommendation: GPT-6 Sol.
who:collab: AI prepares and records; Rufus or the designated content operator supplies experiential facts, operates the real workflow, or participates in a walkthrough. The label is an execution mode, not a claim that another team member has accepted an assignment.
who:rufus: the repository's human-decision label. Rufus is assigned the pilot and adoption decisions. AI may prepare recommendations but cannot substitute for his choice.

Astra is optional for difficult synthesis or workflow/pilot tradeoffs, especially when sources conflict or a proposed improvement affects several processes. It is not required for every step and cannot resolve missing facts or make human editorial decisions. This is a task-specific recommendation, not a measured benchmark on this project. Official model guidance distinguishes Astra for the hardest work from Sol for strong reasoning on demanding tasks. No model setting is changed by this handoff.

Model recommendations are also queryable labels: model:gpt-6-sol on the five AI tasks and the AI-assisted portions of the two collaborative tasks; model-escalation:gpt-6-astra on diagnosis, process design, pilot comparison, and evaluation where difficult judgment may warrant escalation. Human-decision tasks have no model label. These labels are recommendations, not an automatic model switch or authorization for AI to make a human decision.

Working sequence

The dependency graph in Beads is authoritative. In outline: AI evidence audit → collaborative walkthrough → AI diagnosis, minimal process, and pilot package → Rufus's pilot decision → collaborative trial → AI evaluation → Rufus's adoption decision. The five AI steps and two collaborative steps are bounded in their Bead descriptions; the two human decisions are deliberately grouped around concrete proposals and results.

The first ready task is lip-93f.2.1. Start with its evidence matrix and bundle no more than five factual questions for the walkthrough. Continue through ready AI tasks without asking for approval after every document edit. When the next task needs human input, present the prepared material and specific questions together. Do not claim a human decision is complete from elapsed time or an AI recommendation.

bd ready --label handoff:content-pipeline --label who:ai
bd list --label handoff:content-pipeline
bd show lip-93f.2.1

Use the same ready query with who:collab or who:rufus to find the next interactive checkpoint. Claim a task when actually starting it; close it only when its deliverable and acceptance criteria are met. If an evidence gap prevents completion, state the exact gap and which independent work remains possible.

Artifacts and source-of-truth boundaries

projects/2602-content-delivery-pipeline.md: durable framing, compact issue tree, decisions, and agreed direction.
docs/plans/content-pipeline-evidence.md: dated evidence and collaborative walkthrough, including explicit unknowns.
docs/plans/content-pipeline-working-process.md: to be created by the process task; minimal release brief, stage/handoff/review rules, and the proposed pilot package.
docs/plans/content-pipeline-trial-results.md: to be created during the approved trial; actual outputs, effort/waiting/rework observations, evaluation and limitations.
Existing comms issues: operational release work. Reuse them; do not copy the live publication queue into planning Markdown or create duplicate delivery tasks.

The user has authorized this planning and handoff, not a specific public release or tooling implementation. The pilot decision must identify the exact scope, owners and permitted actions before execution. A preparation-only trial must be described as such, not as evidence of a successful public launch. Do not contact colleagues, publish, change account settings, move media, or build a destination merely because it appears as an option in the plan.

Read-only research and local planning updates can proceed. Use the relevant Drive/Docs/Sheets skill if inspecting those sources. Record inaccessible sources explicitly; do not infer asset readiness from a checklist alone. Preserve unrelated worktree edits, commit only the completed task's changes, and keep the Beads export consistent without including unrelated changes in the commit.

Copyable handoff prompt

Continue the content pipeline work in /Users/rgrp/src/life-itself/planning. Read briefings/2026-09-23-content-pipeline-handoff.md, the pipeline project, and the existing evidence note. Strategy remains parked. Use Beads label handoff:content-pipeline; start with the next ready who:ai task, currently lip-93f.2.1. Inspect its full specification, claim it, and work to its acceptance criteria. Default to Sol; Astra is optional for unresolved hard synthesis, not a prerequisite. Complete independent research and drafting, then present bundled factual questions or a concrete decision package at the next who:collab/who:rufus checkpoint. Do not publish, contact others, or start unapproved implementation. Preserve source links and unrelated edits, and report completed work plus the next ready checkpoint.