Chiron 10-card Batch-32 Forward Pass

Built 2026-07-07, reconciled 2026-07-08. Companion: index.md. Master index: indexes-board. Live tracker: chiron-video-tracker.

RECONCILED 2026-07-08: the batch-32 request is being fulfilled by a peer Claude Code session (13c5a020-66c4-4a10-8c12-e692b0a327df, opus-4-7) that is autonomously discovering + rendering the full chiron intuition backlog under the sourceaudio-AUTO-*-10card naming pattern. Ground truth at reconcile time: 40 routes LIVE, 1 rendering right now (the-system-was-built-for-u, PID 13622), 1 staged (forced-to-face-demon). This session did NOT launch a competing orchestrator to avoid Vercel deploy + politeness gate + disk contention. The 34 queue items (parent + discovery + 32 slots) have decisions logged referencing the peer session as fulfiller.
One-at-a-time YouTube-politeness 2×5 rolling window Ladder-not-fanout NEEDS VERIFY voice

The ask, unpacked

Produce 32 new Chiron Intuition 10-card renders, one at a time, staggered so the YouTube politeness gate never trips an IpBlock. The freshest 32 uploads from the @chironintuition channel, filtered against the 16 already shipped in _batch16.json and the 32 live routes under public/chiron-sourceaudio-*-10card.

Hard gates in force

GateWhat it forbids
feedback_one_at_a_time_no_bulk_completeBatching more than one render into a single session pass. Ladder pulls one, ships one, moves on.
feedback_youtube_fetch_politeness_hard_gateRaw yt-dlp / youtube-transcript-api calls. Every fetch routes through ~/.claude/skills/_shared/yt_polite.py: serialized via flock, spaced 45-90s with random jitter, rolling budget cap. PreToolUse hook enforces this.
feedback_chiron_tencard_2x5_rolling_window_hard_gateAny layout that is not the 10-slot 2×5 rolling FIFO (drop-oldest-5 on 11th distinct card).
feedback_chiron_tracker_strict_voice_and_render_rulesMarking a _sourceaudio render as creator voice without a listen-check. All 32 land as NEEDS VERIFY until proven.

The ladder (queue snapshot 2026-07-07)

IDRoleStatus
q-2026-07-07-083fd4PARENT: batch-32 forward-pass. Full plan + hard-gate roll-up.queued
q-2026-07-07-5264deSTEP 1: discovery. Runs fetch_channel_top.py --seed https://youtu.be/6rZVAdKBH9I --order recent --top 48 --slug chironintuition-batch32 through the polite gate, filters, emits _batch32.json.queued
q-2026-07-07-2438b9 … +31Render slots 1/32 through 32/32. Each pulls the first status=queued entry from _batch32.json, renders via chiron-tencard-reading pipeline, ships via /ship gate.queued (executes after discovery lands _batch32.json)

Slot map

Pipeline for each render (per slot)

  1. Read next status=queued entry from _batch32.json. Cards + title + slug already resolved by discovery.
  2. Download source audio for the YouTube ID through yt_polite.py (45-90s stagger baked in).
  3. Transcribe locally (Whisper), align with card pull-order beats, run source_audio_render.py --max-slots 10.
  4. Generate the reading PDF (reading_pdf.py) + card-guide PDF (teaching_pdf.py).
  5. Ship page to public/chiron-sourceaudio-<slug>-10card, run the /ship gate (kid FAB + sibling .md + Vercel 200), verify.
  6. Mark _batch32.json entry status=done. Update chiron-video-tracker.

How to watch the ladder

Ladder status is written by ~/.claude/skills/no-quit-shift/queue.py status. As slots progress through the shift states, this page's grid updates on the next redeploy. The tracker at chiron-video-tracker reflects each completed render inline within the tencard-count section.

Ship gate

Every completed render must exit the /ship script green: one real kid button, sibling .md, Vercel deploy, HTTP 200 on every route, proof block. That is the completion contract for each of the 32.


Part of drwu-htmls.vercel.app. Sibling text version: chiron-batch32-plan/index.md.