Slide archetypes land in the corpus — a vocabulary for the narratives every pitch deck reaches for

Why Care?

Every fundraise deck reaches for the same dozen-or-so narratives — the problem, the moment, the team, the ask — but every team calls them something slightly different. "Why-now" in one deck is "the inflection point" in another and "macro tailwinds" in a third. The slide is the same beat; the slug isn't.

DidiDecks' job is to recognize the beat regardless of slug, then re-render it in the engagement's design system. That requires a canonical vocabulary the system can map noisy real-world slide names back to. This commit installs that vocabulary as a first draft — 27 archetypes covering the two customer shapes most of our engagements take: startups pitching VCs, and VCs pitching LPs.

What's New?

A new folder lives at corpus/slide-archetypes/. It contains 27 markdown files, one per archetype, each carrying a YAML schema:

• default_slug — the canonical filename and identity

• slug_variants — names the same beat goes by in the wild (why-now → the-moment, inflection-point, macro-tailwinds, ...)

• description — what the slide is actually about, in one sentence

• tags — Startup-Pitch-Narratives, VC-Pitch-Narratives, or both

• date_created / date_updated — ISO dates, for living-document hygiene

• often_before_slides / often_after_slides — a soft sequencing graph, referencing other archetypes by their default_slug

The tag distribution shook out:

• Startup-only (10) — problem, solution, market-size, product-demo, business-model, competition, traction, roadmap, ask, use-of-proceeds

• VC-only (6) — fund-thesis, portfolio-construction, track-record, recent-exits, pipeline, lp-terms

• Both tags (11) — why-now, regulatory-tailwinds, positioning, core-team, extended-team, media-mentions, references, who-is-in, lighthouse-customers, thought-leadership, blue-ocean-strategy

The eleven dual-tagged archetypes are the most interesting category — these are the beats that show up regardless of who's pitching whom. Thought-leadership matters whether the founder writes a substack or the GP writes one. Lighthouse customers and anchor LPs are the same trust-by-association move. Blue-ocean-strategy — the Strategy Canvas with its ERIC (Eliminate / Reduce / Raise / Create) framing — repositions a startup against incumbents or a fund against peer funds with the same visual logic.

How It Works

The schema is designed for two concrete things:

Slug matching. When a real deck arrives with a slide titled "The Moment," DidiDecks looks it up in the variants list and resolves it to why-now. The rendered deck then uses the canonical identity, but the original slug is preserved so the variant catalog grows organically. Today this is a catch-all array; eventually a variant might earn its own nuance file alongside the default.

Soft sequencing. Each archetype names slides that often come before and after it. This isn't a hard order — it's a graph of pull. The startup happy-path flows problem → why-now → solution → market-size → product-demo → business-model → competition → positioning → traction → core-team → extended-team → roadmap → ask → use-of-proceeds. The VC happy-path flows fund-thesis → why-now → positioning → track-record → recent-exits → portfolio-construction → pipeline → core-team → extended-team → lp-terms. The graph encodes both without forcing either.

Distinctions the schema deliberately holds:

• problem is about pain; why-now is about timing; market-size is about scale. They're separate beats and they get separate files.

• media-mentions is press about you. thought-leadership is content by you. Different credibility moves.

• references is who'll take a diligence call; who-is-in is who's already committed to the live raise. Different social-proof moves.

• core-team is who runs the thing; extended-team is the credibility ring around it.

Under the Hood

Each file is small — frontmatter plus a short prose paragraph describing the archetype's role and how it differs from its neighbors. The body paragraph is where future nuance will accumulate; today it carries just enough context that a downstream tool (or a downstream human) can disambiguate adjacent archetypes without re-reading the whole catalog.

The corpus submodule is on development at 33b32cf. The parent dididecks-ai pointer bump in this same commit records that tip.

What's Next

• Wire the archetype catalog into /api/slide-decompose so a parsed real-deck slide can be matched to its canonical archetype.

• Render the sequencing graph as a visual aid — a mermaid graph in the corpus README, or a per-tag flow diagram.

• Start an archetypes/variants/ companion folder for the nuance files the schema's catch-all variants array hints at — when a variant turns out to carry a meaningfully different rendering, it earns its own file.

• Watch for archetypes the first two client engagements (calmstorm-decks, reach-edu-hub) hit that aren't in the seed set, and grow the catalog from real decks rather than from theory.