Onboarding Vision

Karpathy-informed onboarding — knowledge-first, not account-first.

The Karpathy LLM wiki signal: the most respected ML researcher chose flat markdown in a git repo over a blog, book, or course. This validated the format and created demand with no supply. wikihub fills the gap — but the onboarding should feel like "publish your knowledge" not "create an account on a platform."

Current flow (generic SaaS)

Proposed flow (knowledge-first)

Arrival: See beautiful rendered wikis (official wiki + featured content) → "I want mine to look like that"

Three entry points:

"Drop your files" — drag a folder of .md files, get a wiki instantly. Account created behind the scenes if needed.
"Start from a template" — pick the Karpathy skeleton (schema.md, index.md, log.md, wiki/, raw/) or a blank wiki.
"Connect your vault" — git remote add wikihub ... + git push for Obsidian users.

Immediate payoff: Content is live and rendered within 60 seconds of arriving. Personal wiki exists, index.md is your profile. You're not "setting up an account," you're publishing.

Discovery comes after creation: Explore, star, fork happen once there's content in the system. Not the entry point.

Concrete implications

Landing page: Show a live preview — "here's what your wiki will look like" with a dogfood wiki rendered as the example. CTA is "publish" not "sign up."
Signup: Nearly invisible. Drag-and-drop or git push creates the account as a side effect. For web: username + optional email, done, you're in the editor. For API: already one POST.
Post-signup: Land in personal wiki's index.md editor, not an empty dashboard. Template pre-filled with Karpathy-style skeleton. Message is "write something" not "configure your account."
/explore: Where the official wiki earns its keep. New users who aren't ready to publish yet browse real content — curated picks, most-starred, official docs. The "what is this place" moment.

Why this matters

The audience (Karpathy-gist wave, Obsidian vault owners, ML researchers) already has content. They don't need to be convinced to write — they need a place to put what they've already written. The onboarding should be a funnel from "I have files" to "they're live" with minimal friction. Account setup, ACL configuration, and social features are all things that happen after the first publish, not before.

See also: Core Beliefs for the foundational philosophy, The Librarian for the AI vision.