The modern creative tool stack: a map
Week 2 gave you the matrix — one idea fanned into fifty tagged assets. The obvious next question is: who makes fifty assets a week? This week, AI does. But only if you stop collecting tools and start building a line.
Your creative stack is not a list of clever AI products — it's a pipeline of eight jobs, and the value lives in the line connecting them, not in any single tool you swap in or out.
On Day 10 you did the arithmetic that should still be making you uneasy: one concept × three personas × three angles × two treatments × three formats is dozens of distinct, genome-tagged assets — and that's one concept, in one week. Hand-produce that at agency rates and you'll spend €40k and six weeks to ship a batch that fatigues in two. The matrix only pays off if you can manufacture it cheaply and fast. That's what Week 3 is about. And it starts not with a tool, but with a map.
1Map the jobs, not the brands
The trap with AI creative tooling is that the product names change every quarter. In the time it takes to read this course, a flagship video model will get discontinued (Sora is sunsetting through 2026), a new image model will top the blind-test leaderboards (Nano Banana Pro did exactly that), and three "AI UGC" startups you've never heard of will launch. If you anchor on brands, your knowledge has a shelf life of about ninety days.
So anchor on jobs instead. There are eight, and they have stayed structurally stable even as the products underneath churn. Each is a station on the production line. Learn the station — what it does, what makes a good one, where it hands off to the next — and you can swap the specific tool any time the market gives you a better one without rethinking your pipeline.
Read the map top to bottom and it tells the story of the whole course. Stations 1–6 are the production half — the raw-material factory that turns a brief into a tagged batch. Station 7 is Meta's own free GenAI, which we open up tomorrow. Station 8 is the learning half — the creative-intelligence layer that reads what won and writes the next brief. You'll notice the engine is already implied here: the bottom of the stack loops back to the top. Hold that thought; it's the whole of Week 4.
2The selection criteria that survive tool churn
Once you think in jobs, "which tool do I buy?" becomes a sane question instead of a doom-scroll through Product Hunt. For every station, ask the same five things — and notice that "is it the most impressive demo?" isn't one of them.
Worked example. Say you need fifty image variants for a skincare matrix — five concepts, two personas, plus background and angle variation. A pure-aesthetic hero model might cost roughly €0.30 an image and need manual brand-fixing on each: fifty images is €15 plus an hour of cleanup. A bulk-optimised model at €0.04 an image, fed a locked brand style reference, lands the same fifty for €2 and hands off clean. Same station, same job — but the second choice is built for the line and the first is built for the demo reel. Multiply that decision across eight stations and across a year of weekly batches and the gap is your entire production budget.
3A stack is a pipeline, not a shelf
Here's the part most people miss. The point of the map isn't to own a tool in every row — it's to make the rows connect. A concept from station 1 becomes a script in station 2, which drives a talking-head in station 5 and an image in station 3, both of which get cut, captioned and resized to every placement in station 6, then tagged with the genome from Day 4 and shipped. Value is created in the handoffs. A brilliant image model that feeds nothing downstream is worth less than a mediocre one wired into a working line.
This is why "which is the best video generator?" is the wrong opening question. The right one is: what is my line, and which tool sits cleanest at each station? A solo founder might run five tools across eight stations — one LLM doubling for ideation and copy, one image model, one AI-UGC tool, one editor, plus Meta-native — and that's a complete, compounding pipeline. A team chasing the "best" tool in every category often ends up with twelve subscriptions and no batch shipped, because nothing connects. Half the job, as ever in this course, is restraint.
A professional kitchen isn't one genius with a magic knife — it's a brigade: a station for prep, one for the grill, one for sauce, one for plating, each handing off to the next. No single station makes the meal; the line does. A new chef who buys the sharpest knives for every station but never builds the line still can't serve dinner. Your eight tool categories are eight stations. The diner only ever tastes the line — and so does Meta's auction.
Your stack shouldn't live in your head or your browser tabs — it lives in one row of your production tracker, next to the genome columns from Day 4. One named owner (or tool) per station, one default tool, one fallback, and the disclosure flag set for anything synthetic (compliance is coming on Day 12 and Day 14). This is the SOP page a new hire reads to run the line on day one.
They become tool collectors. They sign up for nine shiny generators, watch the launch demos, and confuse a full toolbox with a working factory — meanwhile still hand-making the one thing AI could batch, and never shipping a tagged batch at all. The other half of the same mistake: paying a human (or a premium tool) to do a mechanical job — manually resizing a hero asset into eight placements when one exporter does all eight in a click. Your edge is the opposite reflex: name the eight jobs, wire one tool per job into a line, automate the mechanical, and judge the stack by one number — tagged variants shipped per week. Tools don't compound. A pipeline does.
Today's recap — 30 seconds
- The creative stack is eight jobs, not a brand list: ideation, copy, image, video, AI UGC, editing/assembly, Meta-native GenAI, creative intelligence.
- Anchor on the station, not the product — names churn every quarter; the jobs don't. Swap tools freely within a station.
- Choose tools on five durable criteria: fit-for-job, on-brand control, batch/API, cost-per-asset, clean handoff — never "best demo."
- Value lives in the line, not any single tool. A connected five-tool pipeline beats twelve disconnected subscriptions.
- Stations 1–6 are this week's production half; station 7 (Meta-native) is Day 12; station 8 (intelligence) is the Week 4 learning layer that closes the loop.
- Judge the whole stack by one metric: tagged variants shipped per week (the Day 4 genome makes them learnable).