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-W38-Sunkind longformweek W38date 2026-09-20campaign longform-youtubepillar multimediabeat asset videoduration 226.0sground whitescenes 10

Checklist the per-video bar — engine/sim

100.0/100
plain languagevo coverageno dead airuniquenesscaption fitcompletenesscleanliness
quantitative quality · weights learn from your reviews (engine.sim.memory review longform-W38-Sun good|bad)
✓ all static checks pass — one-focal/scene · tier-by-beat · one-track caption · colorway · cast+shape correct · no banned/fabricated. (audio + visual tiers verify on the rendered finals — Phase 2)

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 10 scenes · 226.0s · comp_id + rendered still + tier + the script

#Layer (comp_id · still · tier)BeatTimecodeMotionLogoAudioVO / on-screen / caption
1s1
matches intent
shared field
signature-3d
open0–25.5sspatial-parallaxicon·wireframe♪ bed_in
This week is about one thing: stop guessing where your money goes and start reading the numbers like a meter. Today we go deep on analytics, the part most owners skip. By the end you will know how to wire your own data into a model and let it tell you, in plain words, what is working and what is wasting your cash. No dashboards you never open. Just answers.
on-screen: Stop guessing, start printing
expected on screen: white ground · octa hero in the shared Signal Field · Lumen leads · node-graph · spatial-parallax · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual node-graphshape octaground whitetreatment wireframemotion spatial-parallaxpower summoninstrument summon→Stop guessing, start printing
2s2
matches intent
NumberedList
template
teach25.5–47.1skinetic-buildicon·color♪ node_lock
Step one is to stop trusting the pretty summary screen. Install a plain tag, like Google Analytics 4 or a self-hosted tool like Plausible, and send every important action as a named event: page seen, form sent, cart started, sale made. Name them like you talk. That raw stream is yours. Everything else we do today reads from it.
on-screen: Own the raw event stream
expected on screen: white ground · a NumberedList panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·color logo · caption bottom-left
spec (the prompt): comp_id NumberedListvisual node-graphshape octaground whitetreatment colormotion kinetic-buildpower morphinstrument morph+laser→Own the raw event streamcurate items, nodes
3s3
matches intent
ChecklistCard
template
teach47.1–70.1skinetic-buildicon·color♪ node_lock
Now connect it to a model. Wire an MCP server into the GA4 Data API, hand the agent your property ID, and ask it for the last thirty days of those events. The model pulls the real rows itself instead of you exporting a spreadsheet at midnight. You ask a question in English, it queries your account, it answers from your actual data.
on-screen: Pull it with an MCP
expected on screen: white ground · a ChecklistCard panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·color logo · caption bottom-left
spec (the prompt): comp_id ChecklistCardvisual node-graphshape octaground whitetreatment colormotion kinetic-buildpower morphinstrument morph+laser→Pull it with an MCPcurate items, nodes
4s4
matches intent
LegacyCard
template
teach70.1–93.89999999999999skinetic-buildicon·color♪ node_lock
Pick a single number that means money for each page. For a booking page it is bookings. For a shop it is sales. Not visits, not time on site, the thing that pays you. Tell the agent that number is the goal. From here on, every report it writes is ranked by that, so you stop staring at numbers that feel busy but mean nothing.
on-screen: Name one number per page
expected on screen: white ground · a LegacyCard panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·color logo · caption bottom-left
spec (the prompt): comp_id LegacyCardvisual node-graphshape octaground whitetreatment colormotion kinetic-buildpower morphinstrument morph+laser→Name one number per pagecurate nodes
5s5
matches intent
CodeWindow
template
teach93.9–116.9skinetic-buildicon·color♪ node_lock
Ask the model to build a funnel from your events: saw the page, started, finished. It will show you the exact step where people leave. Most of the time it is one step, and it is dumb, like a form that asks for a phone number nobody wants to give. The data points at the leak. You go fix that one thing first.
on-screen: Follow the drop-off
expected on screen: white ground · a CodeWindow panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·color logo · caption bottom-left
spec (the prompt): comp_id CodeWindowvisual node-graphshape octaground whitetreatment colormotion kinetic-buildpower morphinstrument morph+laser→Follow the drop-offcurate codeLines, nodes, windowTitle
6s6
matches intent
BuildLog
template
teach116.9–140.3skinetic-buildicon·color♪ node_lock
Add simple tags to your links, the ones that say utm source and utm medium, so every visitor carries a label for where they came from. Then ask the agent to group sales by those labels. Now you can see that the email list pays and the paid post does not, in real numbers, not a hunch from the last person you talked to.
on-screen: Tag where they came from
expected on screen: white ground · a BuildLog panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·color logo · caption bottom-left
spec (the prompt): comp_id BuildLogvisual node-graphshape octaground whitetreatment colormotion kinetic-buildpower morphinstrument morph+laser→Tag where they came fromcurate lines, nodes
7s7
matches intent
StepFlow
template
teach140.3–162.60000000000002skinetic-buildicon·color♪ node_lock
Here is the move people miss. Feed the agent the funnel, the sources, and your one money number, then ask it to write a short plain-language read: what changed, why, and the one thing to do next. The model is good at turning a wall of rows into three honest sentences. You read those over coffee instead of squinting at charts.
on-screen: Let the model write the read
expected on screen: white ground · a StepFlow panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·color logo · caption bottom-left
spec (the prompt): comp_id StepFlowvisual node-graphshape octaground whitetreatment colormotion kinetic-buildpower morphinstrument morph+laser→Let the model write the readcurate nodes, steps
8s8
matches intent
AnnotatedDiagram
template
teach162.6–182.79999999999998skinetic-buildicon·color♪ node_lock
Last, make it repeat. Put that whole chain on a timer, a simple cron job that wakes the agent every Monday, pulls the week, and drops the read in your inbox. You built the report once. Now it writes itself forever. That is the difference between a tool you bought and a habit that runs.
on-screen: Run it on a schedule
expected on screen: white ground · a AnnotatedDiagram panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·color logo · caption bottom-left
spec (the prompt): comp_id AnnotatedDiagramvisual node-graphshape octaground whitetreatment colormotion kinetic-buildpower morphinstrument morph+laser→Run it on a schedulecurate callouts, nodes
9s9
matches intent
RankList
template
proof182.8–204.4sreceipts-counticon·color♪ node_lock
We do not just say this. The studio runs on the same loop. Our whole content calendar is built from one source file, and the same kind of agent reads what each piece does and tells us where to spend next. You can watch it run in public, the work and the numbers, at office.temerarii.xyz. No deck, no maybe.
on-screen: We read our own meter first
expected on screen: white ground · a RankList panel over a dimmed Signal Field · Lumen leads · receipts · receipts-count · icon·color logo · caption bottom-left
spec (the prompt): comp_id RankListvisual receiptsshape octaground whitetreatment colormotion receipts-countpower receiptsinstrument spotlight→We read our own meter firstcurate rows, statsLabels
10s10
matches intent
shared field
signature-3d
resolve204.4–226.0scoalescenceicon·wireframe♪ bed_out
So the takeaway is simple. Own your raw events, name the one number that is money, let a model read it for you, and put it on a timer. You stop guessing and start watching the meter tick. Want to see the whole thing assembled and running. It is live at office.temerarii.xyz. Start with one page and one number.
on-screen: Build the meter once
expected on screen: white ground · octa hero in the shared Signal Field · Lumen leads · coalescence · coalescence · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual coalescenceshape octaground whitetreatment wireframemotion coalescencepower coalescenceinstrument coalescence→Build the meter once

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)

youtubeWire your own data into a model and let it read your numbers in plain words Most owners skip analytics or trust a pretty summary screen. This is how to own your raw data, name the one number that means money, and let a model tell you, in plain words, what is working and what is wasting cash. The method: - Own the raw event stream. Install a plain tag like Google Analytics 4 or a self-hosted tool like Plausible, and send every important action as a named event: page seen, form sent, cart started, sale made. That raw stream is yours. - Pull it with an MCP. Wire an MCP server into the GA4 Data API, hand the agent your property ID, and ask for the last thirty days. You ask in English; it queries your account and answers from your own rows. - Name one number per page that means money, not visits. Tell the agent that number is the goal, so every report is ranked by it. - Follow the drop-off. Ask the model to build a funnel from your events and show the exact step where people leave. Usually it is one dumb step. Fix that first. - Tag where they came from with UTM source and medium, then group sales by those labels. Now you see which channels pay in real numbers. - Let the model write the read. Feed it the funnel, the sources, and your money number, then ask for three honest sentences: what changed, why, and the one thing to do next. - Run it on a schedule. Put the whole chain on a timer that wakes the agent every Monday and drops the read in your inbox. You built the report once; now it writes itself. The honest proof: the studio runs on the same loop. Our whole calendar is built from one source file, and the same kind of agent reads what each piece does and tells us where to spend next. Own your raw events, name the one number that is money, let a model read it, and put it on a timer. Start with one page and one number. It is live at office.temerarii.xyz. Keywords: web analytics, GA4 Data API, MCP, conversion funnel, UTM tracking, automated reporting.

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