// pillar 01 β /pipeline
π‘ Idea Pipeline
Most people fall in love with one idea. The machine dates two thousand at a time and only commits when the math clears.
Pillar 01 β Idea Pipeline
π‘ The market math runs before the code
β click a feature β the window follows
$ generate --buckets 24 --count 2000
β¦generated 2,000 Β· deduped 312 Β· knockouts 941
$ score --system vv-saas-g1 --passes 3
β 38 advance β deep research queuedβ
// the breakdown
Every module, all the way down.
pipeline.01
Idea Engine
Volume first. Feelings never.
One run writes thousands of venture ideas across 24 industry buckets β legal ops, logistics, pet care, restaurants, elder care, trades β each with a name, problem, buyer, pricing model, and why-now baked in at birth. Then the gauntlet: an instant knockout pass kills the structurally doomed, a condensed pre-filter scores everything cheaply, and the survivors get full triple-pass scoring against a configurable viability formula.
Scoring systems are swappable and yours to define β a general SaaS formula, a paid-ads-only variant that weights CPA reality, a profit-target build. Old runs re-score under new formulas without regenerating a single idea, so when your thesis sharpens, your whole history re-ranks overnight.
Batch directives ride along from the very first token: tell a run "paid ads only, self-serve checkout, no sales calls" and that constraint follows its ideas through research, briefs, validation, the build, and the campaigns β not just generation.
Throughput
2,000 ideas / run
Scoring
3-pass, formula-swappable
Dedupe
cross-run, canonical-name
Directives
carried to every module
pipeline.02
Bring your own
Your shower thought gets the same gauntlet.
Type an idea straight in β a name and an optional line of description β or upload a whole CSV with any columns you have. Either way it enters the exact machinery generated ideas get: the same dedupe, the same base-level research, the same scoring passes, ranked into the same pool against everything the machine has ever produced.
That last part is the point: your pet idea doesn't get a participation trophy. It gets a number, next to two thousand rivals, and the number decides.
Input
typed or CSV/JSON
Mapping
auto, any headers
Treatment
identical to generated
Verdict
a rank, not an opinion
pipeline.03
Deep Research
Winners earn a dossier.
Ideas that clear the bar get a full research pass: competitor landscape with real named rivals, ICP mapping down to the trigger moment, distribution channels with viability calls, and the why-now case. Every section is drillable β hit Dig Deeper on the competitor section and a focused pass goes three levels further on just that question.
Then the loop closes: a research-informed re-score reads the report's current content β including everything the deeper digs surfaced β and re-rates the idea. A 74 that researches into a 77 advances; a 74 that researches into a 58 just saved you four weeks of building the wrong thing.
Sections
drillable, individually
Re-score
reads the current report
Output
advance / pause / kill
Compare
deep scores across ideas
pipeline.04
Concept Briefs
One page. One decision.
Before a single file gets built, the idea is forced onto one page: the problem in plain words, the solution, why now, the pricing model, the distribution plan, and the exact test that will prove or kill it. Briefs are generated from the idea's research so they inherit real competitor and channel context, and they end in a decision β not a vibe.
The brief follows the idea everywhere as its card: open it from the board, the builder, the campaign planner β same page, same truth.
Length
one page, enforced
Source
research-grounded
Ends in
a decision
Lives
on the idea, forever