

Controls the LIVE engine for the whole team - the desktop app's Settings edits these same values (whichever saved last wins). Applies within a minute of saving.
The AI write-up runs a few times a day, not on every refresh (the main Anthropic-cost lever). Between runs it shows the last analysis with its time. The action worklist is always live.
Scan cadence. Longer intervals ease the Meta/Triple Whale API load (the shared rate budget the buying team needs for bulk edits). Safe to adjust anytime.
Last-rung rename is live under 6.1.1 rules: an AD named TESTING flips to SCALING at the last spend rung (about $950 on the 0–15% band). The SET stays TESTING while a live ad still says testing. This leftover toggle is the older hygiene rename (name contains testing). It swaps the word, then re-reads it. Nothing else about the ad set changes.
Replicates your post-ID SOP through the API and verifies each change. Partnership ads it can't post-ID are flagged, not touched. Start the cap low, check the first batch in Meta, then raise it.
AUTONOMOUS CREATIVE TESTING. Buff Jr. already runs 6.1.1 rules on 0089 and BAU (both stay in scope). No max kills or doubles per run (0 = unlimited). Dollar caps $500 / $1,000 stay. These leftover fields do not cap the live Buff Jr. lanes.
Every rule that fires a row in this app, in one place. All money decisions are deterministic API-data rules — no AI in any kill, scale, or winner call. AI only writes summaries, advisory verdicts, and the chat. Creative testing runs 6.1.1 rules. Every BUFF JR. ACTION LOG row cites the exact rule that fired via its blue ⓘ why button.
| Source | Owns | Attribution / window | Refresh |
|---|---|---|---|
| Triple Attribution | Kill/2X cells: CPA, ROAS, NC ROAS. TW NVP is orders-share evidence, never the band. | Ad-level kill/tag is 1-day; 2X margins are lifetime | near-real-time (15-min scans) |
| Northbeam | Kill / 2X / revival band (visits %), plus CPO / ROAS / NC ROAS cells | 7-day click, clicks-only | ~hourly (Northbeam's own schedule) |
| Meta | Live spend, status, budgets, change history — verifies check-offs | live / today | ~15 min |
Winning on BOTH platforms at once, over the last 7 days: Northbeam $500+ spend, 5+ purchases, CPO ≤ $45, ROAS ≥ 0.8 AND Triple Whale $500+ spend with ROAS ≥ 1.5 or CPA ≤ $35. The $500 floors exist so one hot day can't mint a mega winner. Rescanned every 15 min.
Precedence LD1 > LD3 > LD2. Tag the ad LAUNCHDAY1/2/3 in Meta after acting and it drops off the next scan.
CDV (clicks + modeled views) needs time to mature: today's CDV is never used, 3+ days back is real. CDV lens revival is retired 2026-08-15. Live 6.1.1 revival is the 8-day NB-NVP watch. CDVMAXXING is a Testing-inactive harvest, not a CWIN dest.
✓ Done → the engine checks Meta on the next refresh: kills clear only when Meta shows the ad set PAUSED; scale-downs clear when the budget is cut ~30%. If Meta shows no change, the row pops back with a warning. ✗ Won't do → moves to the OPTED OUT card and the AI re-reviews it in the next report.
Fastest way to interrogate any number: the ⓘ dot on it, or ask the 🦬 AI chat (bottom right) - it runs on this same rule set plus your live data.
