Back Office · office.temerarii.xyz
One asset, all the way in — composition, the wireframe + storyboard, the output format stack, and the template, all read from the SAME content-index record. The expected output matches what /media surfaces for this post.
post longform-W45-Satkind longformweek W45date 2026-11-14campaign longform-youtubepillar socialbeat —asset videoduration 149.9sground blackscenes 8

Checklist the per-video bar — engine/sim

98.4/100
plain languagevo coverageno dead airuniquenesscaption fitcompletenesscleanliness
quantitative quality · weights learn from your reviews (engine.sim.memory review longform-W45-Sat good|bad)
⚠ 1 flag(s) — not yet ship-ready: custom_element_dup · see docs/strategy/VIDEO-CHECKLIST.md

Composition comp · template family · expected output

composition LongFormChaptersfamily / template LongFormChapters
9:16 Reelpending1:1 Squarepending16:9 Widepending9:16 4Kpending1:1 4Kpending16:9 4KpendingGIF (SMS)pending
render pending — silent master not yet on disk
expected output: 0/7 rendered — same matrix the /media preview surfaces for this asset.

Composition layer × scene 8 scenes · 149.9s · comp_id + rendered still + tier + the script

#Layer (comp_id · still · tier)BeatTimecodeMotionLogoAudioVO / on-screen / caption
1s1
matches intent
shared field
signature-3d		open0–18.8sspatial-parallaxicon·ember-fill♪ bed_in
Posting is half the engine. The other half is what happens after, in the comments and the DMs. Today is community, the quiet work that turns a feed into a relationship. By the end you will know how to let a machine help you stay present without living in your phone.
on-screen: The part everyone forgets
expected on screen: black ground · torus hero in the shared Signal Field · Nuntius leads · node-graph · spatial-parallax · icon·ember-fill logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual node-graphshape torusground blacktreatment ember-fillmotion spatial-parallaxpower summoninstrument summon→The part everyone forgets
2s2
matches intent
NumberedList
template		teach18.8–37.6skinetic-buildicon·wireframe♪ node_lock
First move. Stop hopping between apps to find comments. Wire the platforms' feeds into the terminal through MCP so the agent collects every new reply and DM in one list. You read one stream instead of five inboxes. Nothing slips through because it was on the app you forgot to open.
on-screen: Pull every reply into one place
expected on screen: black ground · a NumberedList panel over a dimmed Signal Field · Nuntius leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id NumberedListvisual node-graphshape torusground blacktreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Pull every reply into one placecurate items, nodes
3s3
first render · fix pending
ChecklistCard
templatecustom_element_dup		teach37.6–56.400000000000006skinetic-buildicon·wireframe♪ node_lock
Second move. Most replies are noise. Ask the model to read the stream and flag the few that matter, a real question, a buying signal, an angry customer. It sorts the pile so you spend your minutes on the messages that move money or fix trust, not on every thumbs up.
on-screen: Let the model sort the noise
expected on screen: black ground · a ChecklistCard panel over a dimmed Signal Field · Nuntius leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id ChecklistCardvisual node-graphshape torusground blacktreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Let the model sort the noisecurate items, nodes
4s4
matches intent
DiffCard
template		teach56.4–75.2skinetic-buildicon·wireframe♪ node_lock
Third move. For each flagged message, have the model draft a short answer in your voice, using the corpus you fed it earlier. You read it, nudge a word, send it. You stay the human at the door. The machine just removes the blank-page pause that makes replies take all day.
on-screen: Draft the reply, you approve
expected on screen: black ground · a DiffCard panel over a dimmed Signal Field · Nuntius leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id DiffCardvisual node-graphshape torusground blacktreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Draft the reply, you approvecurate lines, nodes
5s5
matches intent
BlueprintGrid
template		teach75.2–94.30000000000001skinetic-buildicon·wireframe♪ node_lock
Fourth move. When three people ask the same thing, that is a post waiting to happen. Have the agent watch for repeated questions and flag them. You turn the answer into next week's content. The community tells you exactly what to make next, if you let the machine listen for the pattern.
on-screen: Catch the same question twice
expected on screen: black ground · a BlueprintGrid panel over a dimmed Signal Field · Nuntius leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id BlueprintGridvisual node-graphshape torusground blacktreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Catch the same question twicecurate nodes, rows
6s6
matches intent
TerminalRun
template		teach94.3–113.8skinetic-buildicon·wireframe♪ node_lock
Fifth move. Log the real conversations into the same record you use for posts. Who asked what, what you promised. Next time they show up, the model reminds you. You remember people, at scale, because the machine holds the memory you cannot. That is how a feed starts to feel like a person.
on-screen: Keep a record of the talk
expected on screen: black ground · a TerminalRun panel over a dimmed Signal Field · Nuntius leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id TerminalRunvisual node-graphshape torusground blacktreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Keep a record of the talkcurate nodes
7s7
matches intent
ComparisonTable
template		proof113.8–130.4sreceipts-counticon·wireframe♪ node_lock
Honest proof. Our own replies run through this loop. One stream, the model flags what matters, drafts a start, we approve. The same community work you just learned keeps The Big T-M talking back in public, with one person present, not a hidden support team.
on-screen: We answer ours this way
expected on screen: black ground · a ComparisonTable panel over a dimmed Signal Field · Nuntius leads · receipts · receipts-count · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id ComparisonTablevisual receiptsshape torusground blacktreatment wireframemotion receipts-countpower receiptsinstrument spotlight→We answer ours this waycurate colA, colB, rows, statsLabels
8s8
matches intent
shared field
signature-3d		resolve130.4–149.9scoalescenceicon·ember-fill♪ bed_out
That is community, the half nobody films. One stream, sorted by the model, drafts you approve, patterns you turn into posts, a memory of every talk. Quiet, human, and real. Watch the full social engine run in public at office dot temerarii dot xyz, then point it at your own comments this week.
on-screen: See the engine at the office
expected on screen: black ground · torus hero in the shared Signal Field · Nuntius leads · coalescence · coalescence · icon·ember-fill logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual coalescenceshape torusground blacktreatment ember-fillmotion coalescencepower coalescenceinstrument coalescence→See the engine at the office

Format stack 1 aspects · same scenes[], re-cropped

16:9
1920×1080
X/Twitter · YouTube · LinkedIn video

Channels 2 destinations

YouTubeBlog

Social captions supplemental published copy · per channel (comp_id level)

youtubeCommunity Management With AI: Sort the Comments, Draft the Replies, Stay Human Posting is half the engine. The other half is what happens after, in the comments and the DMs. In this video we show how to let a machine help you stay present without living in your phone. The moves: - Pull every reply into one place. Wire the platforms' feeds into the terminal through MCP so the agent collects every new reply and DM in one list. You read one stream instead of five inboxes. - Let the model sort the noise. Most replies are noise. Ask the model to flag the few that matter, a real question, a buying signal, an angry customer, so you spend your minutes where they move money or fix trust. - Draft the reply, you approve. For each flagged message the model drafts a short answer in your voice, using the corpus you fed it. You read it, nudge a word, send it. You stay the human at the door. - Catch the same question twice. When three people ask the same thing, that is a post waiting to happen. The agent flags repeated questions and you turn the answer into next week's content. - Keep a record of the talk. Log the real conversations into the same record you use for posts. Next time someone shows up, the model reminds you what you promised. The honest proof: our own replies run through this loop. One stream, the model flags what matters, drafts a start, we approve. One person present, not a hidden support team. The takeaway: one stream, sorted by the model, drafts you approve, patterns you turn into posts, a memory of every talk. Quiet, human, and real. Watch it run at office.temerarii.xyz. Keywords: community management, social media replies, AI agent, MCP, comment triage.

Cross-links every lens is a view on this one record

→ Wireframe · scenes→ Calendar week→ Media (audience preview)→ Storyboard hub→ Day view→ Channel view