diff --git a/README.generators.md b/README.generators.md
index 361eacf60d65e15738b487d8e0349e5cf55e3d6a..eda9bba796f54b29f3eec9fe4e3d62016572c484 100644
--- a/README.generators.md
+++ b/README.generators.md
@@ -1,6 +1,8 @@
 Reference repository generators
 ===============================
 
+FIXME this file is partially outdated
+
 This directory contains the input files and scripts needed for generating:
 
 - The Reference API ie. the JSON files that are served by the Grid'5000 API for describing nodes, network equipments, topology ...
diff --git a/README.md b/README.md
index ac1d3ba62ae7721b645dc0f3f816712efb8256ce..70839ae1e12e9c540bb25299b675c2ac08d9030b 100644
--- a/README.md
+++ b/README.md
@@ -1,56 +1,26 @@
-FIXME update this doc
+'''reference-repository''' is Grid'5000's ''single source of truth'' about sites, clusters, nodes, and network topology.
 
-The reference API data of Grid'5000 is stored in a git repository as JSON files. Those files are organized into hierarchical folders (see data/). The git repository comes with scripts to ease the generation of those API files (see generators/). The generator uses data from input/. The node information (input/grid5000/sites/*/clusters/*/nodes/*.yaml) is fetched using [g5k-checks](https://github.com/grid5000/g5k-checks). The other input files are created manually.
+There are several important parts:
 
-The git repository also includes scripts for generating:
-* The OAR properties ie. the node information that is registered in OAR databases. OAR properties allow user to select resources matching their experiment requirements.
-* The configuration files of the following puppet modules: bindg5kb, conmang5k, dhcpg5k, kadeployg5k and lanpowerg5k.
+* data/ is the set of JSON files that describe Grid'5000. They are exposed by the ''reference API'' (for example: `curl -k https://api.grid5000.fr/sid/sites/nancy/clusters/graoully/nodes/graoully-1.json?pretty`
+* input/ is the set of YAML files used to generate data/. The files in input/ are either:
+    + manually created and edited by admins
+    + generated by g5k-checks (this is the case of the nodes descriptions, in the nodes/ directory of each cluster)
+* validators and generators
+    + to check input/ data for consistency
+    + to generate the configuration of various services (and the list of OAR properties)
 
-Automated tasks are provided to validate and generate data:
+All tasks are executed using '''rake'''. To see the list of tasks, use '''rake -T'''.
 
-```
-rake oar:properties          # Generate oar properties
-
-rake puppet:all              # Launch all puppet generators
-
-rake puppet:bindg5k          # Generate bindg5k configuration
-
-rake puppet:conmang5k        # Generate conmang5k configuration
-
-rake puppet:dhcpg5k          # Generate dhcpg5k configuration
-
-rake puppet:kadeployg5k      # Generate kadeployg5k configuration
-
-rake puppet:kavlang5k        # Generate kavlang5k configuration
-
-rake puppet:lanpowerg5k      # Generate lanpowerg5k configuration
-
-rake reference-api           # Creates json data from inputs
-
-rake validators:homogeneity  # Check homogeneity of clusters
-
-rake validators:schema       # Check input data schema validity
-
-rake wiki:all                # Launch all wiki generators
-
-rake wiki:cpu_parameters     # Generate the media parts for cpu_parameters page
-
-rake wiki:oar_properties     # Generate the media parts for oar_properties page
-
-rake wiki:disk_reservation   # Generate the media parts for disk_reservation page
-
-rake wiki:hardware           # Generate the media parts for hardware page
-
-rake wiki:site_hardware      # Generate the media parts for site hardware pages
-
-rake wiki:site_network       # Generate the media parts for site network pages
-```
+See also:
 
-For more information about generators and validators, please see generators/README.md.
+* https://www.grid5000.fr/mediawiki/index.php/Reference_Repository
+* .gitlab-ci.yml, that automates various checks and the deployment of data files after each commit
+* README.generators.md
 
 # Credentials
 
-all tools should use ~/.grid5000_api.yml. Example:
+all tools that require to authenticate against Grid'5000 use credentials from ~/.grid5000_api.yml. Example:
 ```
  uri: https://api.grid5000.fr/
  username: username