/ r&d sandbox

experiments

tl;dr — prototype agents, spike work, performance benchmarks, throwaway integrations. anything not ready for the production repos but worth checking in for next-time breadcrumbs. each experiment lives in its own dir under experiments/ with its own package.json + readme. when it earns its way out: graduate, archive, or kill.

/01 what goes here

• prototype agents not yet ready for the fleet
• spikes on new features (CRDT, edge runtime, SSO)
• performance benchmarks + load tests
• throwaway integrations + scrapers
• internal tooling experiments
• one-off scripts that earned the right to be checked in
• cross-repo investigations

/02 what doesn't

• production code (use the relevant repo)
• customer-facing features (use the web app / extension / mobile)
• anything destined to ship within 30 days
• cross-repo features (split per repo)

/03 rules of the road

• every experiment lives in its own dir: experiments/YYYY-MM-DD-slug/
• each one has a readme with hypothesis, what we measured, what we learned, decision
• no production credentials; throwaway keys only
• no customer data; synthesize what you need
• time-boxed with an abandon-by date
• public-default mindset even though the repo is private

/04 the graduation path

every experiment finishes with one of four decisions: graduate into an existing repo, spin out a new repo, archive with a why-we-stopped note, or kill + delete. see docs/PROMOTION.md for the full procedure.

/05 live experiments

the live index is at /experiments.

/06 related

• whitelabel-web-app — production web app + api
• whitelabel-extension — chrome extension
• whitelabel-mobile-app — expo mobile app
• whitelabel-agents — agent fleet runner
• whitelabel-status — status page
• whitelabel-shop — storefront
• whitelabel-website — marketing site