Checklist the per-video bar — engine/sim
100.0/100
quantitative quality · weights learn from your reviews (
engine.sim.memory review longform-W31-Thu 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
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 7 scenes · 137.3s · comp_id + rendered still + tier + the script
| # | Layer (comp_id · still · tier) | Beat | Timecode | Motion | Logo | Audio | VO / on-screen / caption |
|---|---|---|---|---|---|---|---|
| 1 |  matches intent shared field signature-3d | open | 0–23.0s | spatial-parallax | icon·ember-fill | ♪ bed_in | Cut busywork, keep revenue. Today it is paid advertising, where money disappears fastest. Most teams set an ad live and forget to check if it pays back. We are going to show you how to let a model read your ad numbers daily and pull the losers before they drain the budget. The busywork is the babysitting. We hand it to a machine. on-screen: Stop feeding ads that lose |
expected on screen: white ground · box hero in the shared Signal Field · Faber leads · node-graph · spatial-parallax · icon·ember-fill logo · caption bottom-left spec (the prompt): comp_id shared Signal Fieldvisual node-graphshape boxground whitetreatment ember-fillmotion spatial-parallaxpower summoninstrument summon→Stop feeding ads that lose | |||||||
| 2 |  matches intent NumberedList template | teach | 23.0–43.5s | kinetic-build | icon·wireframe | ♪ node_lock | Step one. Get an API token from your ad platform, the Meta Marketing API or Google Ads API. That token lets a model pull spend and results without you logging into a dashboard every morning. Wire it into Claude Code through an MCP server so the data comes to the model, not the other way around. on-screen: Connect the ad numbers |
expected on screen: white ground · a NumberedList panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left spec (the prompt): comp_id NumberedListvisual node-graphshape boxground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Connect the ad numberscurate items, nodes | |||||||
| 3 |  matches intent ChecklistCard template | teach | 43.5–62.6s | kinetic-build | icon·wireframe | ♪ node_lock | Step two. Do not judge ads by clicks. Tell the model to divide spend by actual conversions for each ad, so you see cost per booking or cost per sale. Clicks feel good and mean little. A model doing this math every day shows you which ads cost more than they return. on-screen: Measure cost per real result |
expected on screen: white ground · a ChecklistCard panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left spec (the prompt): comp_id ChecklistCardvisual node-graphshape boxground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Measure cost per real resultcurate items, nodes | |||||||
| 4 |  matches intent PipelineMap template | teach | 62.6–82.8s | kinetic-build | icon·wireframe | ♪ node_lock | Step three. Set a clear ceiling. If an ad's cost per result climbs past your limit for two days straight, have the model pause it through the same API. No meeting, no Monday review. The model acts on your rule the moment the rule is broken, so a bad ad cannot quietly eat a week. on-screen: Pause the bleeders automatically |
expected on screen: white ground · a PipelineMap panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left spec (the prompt): comp_id PipelineMapvisual node-graphshape boxground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Pause the bleeders automaticallycurate nodes, stages | |||||||
| 5 |  matches intent BlueprintGrid template | teach | 82.8–101.9s | kinetic-build | icon·wireframe | ♪ node_lock | Step four. Ask the model to find the ad with the lowest cost per result and move spend toward it. This is just steering money to what works. The model does it on the numbers, not on which ad someone on the team happens to like. Feelings stay out of the budget. on-screen: Shift budget to the winner |
expected on screen: white ground · a BlueprintGrid panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left spec (the prompt): comp_id BlueprintGridvisual node-graphshape boxground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Shift budget to the winnercurate nodes, rows | |||||||
| 6 |  matches intent RankList template | proof | 101.9–121.7s | receipts-count | icon·wireframe | ♪ node_lock | We run our own promotion this way and we do not pretend it is glamorous. The model watches our spend, pulls our losers, and feeds our winners. We will not invent a return number to impress you. We will tell you we stopped wasting money the day we let the machine watch the meter. on-screen: We run our own ads honestly |
expected on screen: white ground · a RankList panel over a dimmed Signal Field · Faber leads · receipts · receipts-count · icon·wireframe logo · caption bottom-left spec (the prompt): comp_id RankListvisual receiptsshape boxground whitetreatment wireframemotion receipts-countpower receiptsinstrument spotlight→We run our own ads honestlycurate rows, statsLabels | |||||||
| 7 |  matches intent shared field signature-3d | resolve | 121.7–137.3s | coalescence | icon·ember-fill | ♪ bed_out | So the takeaway. Ads keep revenue when something checks the cost per real result every day and acts fast, and a model is built for exactly that. No babysitting. Go watch the loop in the open at office dot temerarii dot xyz. on-screen: See the meter run live |
expected on screen: white ground · box hero in the shared Signal Field · Faber leads · coalescence · coalescence · icon·ember-fill logo · caption bottom-left spec (the prompt): comp_id shared Signal Fieldvisual coalescenceshape boxground whitetreatment ember-fillmotion coalescencepower coalescenceinstrument coalescence→See the meter run live | |||||||
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)
youtubeLet a model watch your ad spend and pause losing ads automatically
Most teams set an ad live and then forget to check whether it pays back. This walks through how to hand that daily babysitting to a model so a bad ad cannot quietly eat a week of budget.
What you learn, step by step:
- Get an API token from your ad platform (Meta Marketing API or Google Ads API) and wire it into Claude Code through an MCP server, so the numbers come to the model instead of you logging into a dashboard each morning.
- Stop judging ads by clicks. Have the model divide spend by actual conversions for each ad, so you see cost per booking or cost per sale, not vanity clicks.
- Set a ceiling. If an ad's cost per result climbs past your limit for two days running, the model pauses it through the same API. No meeting, no Monday review.
- Ask the model to find the ad with the lowest cost per result and shift budget toward it, on the numbers, not on which ad the team happens to like.
We run our own promotion this way. We will not invent a return number to impress you. The honest point: ads keep revenue when something checks the cost per real result every day and acts fast, and a model is built for exactly that.
See the loop running in the open at office.temerarii.xyz.
Keywords: ad automation, cost per result, Meta Marketing API, Google Ads API, MCP, Claude Code