Custody
over
Autonomous
Commerce
Grawth is an engine that researches angles, renders creatives, verifies them at the edge and stages paid distribution for your catalog, then waits for your hand before a dollar moves.
- 1
Connect the catalog
Point Grawth at your store. Every SKU becomes the white core the machine drills from, priced, costed and dated. - 2
Watch it drill
Agents mine reviews, ad libraries and threads for angles, cut variants in 2K, and reject anything that fails the optical edge guardrails. - 3
Arm the capital
Verified ad sets wait on Meta, PAUSED. Nothing spends until you slide a tranche past the gate, with a ten-second window to void it.
Five-Stage Execution Ledger
Stage 01 of 05
Mine verified buyer language before cutting a single frame.
Grawth ingests your product catalog, reviews, support tickets, and competitor ad libraries, then sorts every claim into three buckets so ungrounded hooks never enter production.
01 · Input Signal
Catalog SKUs, unit COGS, 1,240+ buyer reviews, and live competitor ad library transcripts.
02 · Verification Gate
Three-Bucket Claim Gate separates verified buyer proof from unsupported copy.
03 · Committed Output
Dated angle brief with verified hook moves and unit-checked bundle ladders.
| Primary angle move | Skeptic converts · 91% evidence confidence |
|---|---|
| Supporting buyer proof | 64 review mentions · 11 competitor ads |
| Claim verification | Verified supported · zero medical or fabricated claims |
| Approved bundle ladder | 1-Pack (Save 29%) · 2-Pack (Save 43%) · 3-Pack (Save 50%) |
The cockpit, running
7-chapter guided tour · click anywhere in the window to take control
grawth
Starting engine...
1 / 7 · The Machine Assembles. The catalog core lights first; research, creatives, QA, Meta and capital rings bud outward from it.
Drill what buyers already said. Stop paying a studio to guess.
Manufacturing demand means a studio, a shoot and three weeks per concept. Drilling means mining what customers already wrote in reviews, cutting it into verified creatives in minutes, and paying to test only what passed optical QA.
- 1
Drill what already sells
Angles come from your own reviews, support tickets and competitors' live ads: evidence first, creative second. - 2
Verify at the edge
Every cutout is measured across a 3px perimeter rim for luminance delta and mask overlap. A halo is an automatic rejection. - 3
Keep custody of capital
The engine can stage an ad set. It cannot arm one. Every tranche carries a ceiling, a signature and a ten-second undo window.
| Studio shoot | Grawth pipeline | |
|---|---|---|
| Cost to a tested creative | $1,800 | $14 |
| Time to first verified creative | 21 days | 38 min |
| Variants per angle | 1 to 2 | 3 to 4 |
| Edge QA before spend | Eyeballed | Measured, hashed |
| Who releases capital | Media buyer | You, by hand |
3px Perimeter Halo Verification
Drag the cutout edge contrast against the 3px background rim. Any creative exceeding the 40.0 halo ceiling is rejected before a cent of ad spend is staged.
| Background rim luminance (3px) | 143.0 |
|---|---|
| Cutout edge luminance | 177.5 |
| Meta distribution status | STAGED PAUSED |
Bundle Economics & Margin Guardrail
Every discount and multi-pack saving is checked against unit COGS and compare-at price. An inverted margin ladder halts before a landing page or ad set can compile.
| Bundle tier | Per unit | Buyer saving | Bundle delta |
|---|---|---|---|
| Single Unit · 1 Pack | $38.00 / unit | Save 29% vs $54 | Baseline unit |
| Two-Pack Bundle | $30.40 / unit | Save 43% vs $108 | +20% bundle saving |
| Three-Pack Bundle | $26.60 / unit | Save 50% vs $162 | +30% bundle saving |
Reserve & Extraction · 9 Interactive Plates
Nine interactive proofs of why grounded direction beats synthetic volume.
Arrow keys step stages · Space pauses
Plate 01 · The Residue
Grawth keeps what every test leaves behind.
An ad library only shows the survivor. Grawth keeps every scraped review, rejected hook, halo-failed cut, and killed ad set dated in the vault beneath the winner.
Every product run generates hundreds of review excerpts, hook angles, render cuts, and side tests on a dated timeline.
Questions
- Does Grawth spend money on its own?
- No. It researches, renders, verifies and stages ad sets on Meta in a PAUSED state. Only an operator sliding the capital gate releases a tranche, and every tranche has a ceiling and a ten-second undo.
- Which models does it use?
- Creative renders come from Nano Banana Pro 2K. Every render's receipt records the model, the stage that ran it, and hashes of its prompt, inputs and output.
- What does QA actually check?
- Cutout quality at the edge: rim luminance (at or below 150), edge-to-rim contrast delta (at or below 40.0 halo ceiling) and mask IoU (0.93+). Halos and fringing fail. Failed creatives never reach Meta.
- What will I see in the cockpit?
- A live execution graph of every SKU, angle, render, verdict, ad set and order; a provenance inspector for any of them; the capital gate; and unit economics (MER, contribution, throughput, QA pass ratio). No vanity metrics.
- Which channels does it distribute to?
- Meta first. Distribution is a stage, so other channels slot in behind the same QA and the same gate.
- Can I prove what ran?
- Yes. Receipts are hashed and chained along the lineage, so any order can be walked back through its ad set, QA verdict and render to the research hook that started it.