diff --git a/NEXT.md b/NEXT.md
new file mode 100644
index 000000000..b59e8fa0d
--- /dev/null
+++ b/NEXT.md
@@ -0,0 +1,84 @@
+1.0.0
+
+# setup
+
+```
+lerna boostrap
+npm run next:build
+cd packages/phenomic-docs
+npm start
+```
+
+(`npm run build` for static build)
+
+# what's todo
+
+A fucking lot
+
+- Finish & refacto current packages
+  - phenomic-react & phenomic-webpack as plugins
+  - allow preset (array of plugins)
+- replace main npm scripts workflow for new implementation (break/disable all the things)
+  - add stuff in checklist to not forget if necessary
+- check what's missing from the checklist
+  - do what's missing that we still want (check) + provide migration doc
+    - implement missing thing
+    - remove from old codebase the thing
+  - or release notes for removed stuff (~~delete from list like that~~)
+- review docs & homepage
+- RELEASE!
+
+## checklist
+
+From https://github.com/MoOx/phenomic/issues/925#issuecomment-271502547
+
+Note that stuff can be missing
+
+- [ ] webpack can be boosted with hard source plugin (cache)
+- [ ] internal links from markdown are not doing full page load, but caught and leveraging phenomic optimistic loading (note that “internal” but not unknown links (eg: /project/static-but-not-handled-by-phenomic.html) are not caught. (see `src/_utils/catch-links`)
+- [ ] `src/_utils/error-formatter` transform errors to make those more readable (adjust path + remove useless stuff)
+- [ ] (terminal) logs are created using a single module for coherence (`src/_utils/log` + src/builder/log-formatter)
+- [ ] url + pathname for the website is defined once via the full website url (http://example.com/root/)
+- [ ] offline options to enable service worker (for now via offline-plugin) that auto-register content
+- [ ] scripts in markdown (json) are correctly escaped (`src/_utils/serialize`) to avoid json parsing issues
+- [ ] one util transform all the paths to uri and remove extension like .md/txt/asciidoc etc (`src/_utils/urlify`)
+- [ ] webpack 1 and 2 are supported (but I guess webpack 2 can be the only one - [ ] not a lot of changes anyway)
+- [ ] works on windows (it’s a feature yeah)
+- [ ] src/bin/check-engine check that you use the correct node/npm/yarn version before executing phenomic job (to avoid issue on github etc)
+- [ ] $ phenomic start
+- [ ] $ phenomic build
+- [ ] $ phenomic setup: ask questions to create your config and copy the base theme
+- [ ] some const PHENOMIC_* are injected during the build (src/builder/config.common.js)
+- [ ] sourcemap works out of the box during development, even for static build that use webpack build for node build (🙃)
+- [ ] phenomic check webpack config to avoid error
+- [ ] an “assets” folder can be copied from “content” folder - [ ] not sure this make sense with the new architecture - [ ] but people should be able to reference assets from markdown and that should work “as expected”
+- [ ] phenomic open project in a tab when you “start” and reuse an existing chrome tab if already open (src/builder/openChrome)
+- [ ] postbuild: automatic creation of a CNAME if CNAME option is true, from the config url
+- [ ] postbuild: create a .nojekyll file if “nojekyll” is true (default) (src/builder/post-build
+- [ ] logs show elapsed time by default
+- [ ] dev server show a page saying that JS is required (see src/builder/server)
+- [ ] hot loading for modules AND collection content
+- [ ] webpack logs are limited by default to avoid noize
+- [ ] “/stuff/index.html” works during development (https://github.com/MoOx/phenomic/issues/808)
+- [ ] dev server default page show JS error if one occurs (see src/builder/server)
+- [ ] dev server handle occupied ports - [ ] if default phenomic ports is not free, we should ask user to double check and offer to use another one (like CRA)
+- [ ] react-router scroll mess is handled (see src/client/should-update-scroll.js)
+- [ ] glamor & aphrodite should be handled or VERY easy to enable
+- [ ] react-helmet should be handled (and/or see next.js head feature)
+- [ ] Link handle / & /index.html as a same url for activeState
+- [ ] Link should handle local and external link (so you can reference a dynamic value (eg from front-matter) using Link - [ ] see src/components/Link)
+- [ ] optimistic loading/layout & shit (see src/components/PageContainer) - [ ] big deal here
+- [ ] all the configurations is validated and you get proper warnings if incorrect (see src/configurator)
+- [ ] all options! src/configurator/definitions.js
+- [ ] offer helper to retrieve collections data (I guess it’s done already :) )
+- [ ] base theme offer a package for a recommended eslint config, recommended stylelint config
+- [ ] provide a good default for markdown transformation (see `src/loader-plugin-*`)
+- [ ] easy rss & sitemap
+- [ ] ability to define routes (eg: tags/:tag) with pagination
+- [ ] we should keep or base theme, people seems to like it
+- [ ] good docs!
+
+“Developer features”
+- [ ] npm or yarn can be used (even if npm is slower and less predictable)
+- [ ] project is run on travis for yarn, circle ci for npm and appveyor for windows/npm
+- [ ] project CI should run unit test + build docs + build a new project using the setup and run some end to end test (see e2e-tests)