diff --git a/README.md b/README.md
index dc2f5d1..1d617d6 100644
--- a/README.md
+++ b/README.md
@@ -3,7 +3,8 @@
 This is the repository for 
 [Unjournal](https://www.unjournal.org) evaluations, meta-analysis, and meta-science. 
 
-Our current reports are available at <https://unjournal.github.io>.
+Outputs and reports from here are published at <https://unjournal.github.io>.
+
 
 # How it works
 
@@ -15,7 +16,13 @@ a [Shiny](https://shiny.posit.co) app at
 
 The site can be re-rendered by running `quarto render` from the command line,
 or `quarto::quarto_render()` from within R, and pushing changes to the `main` branch.
+The rendered HTML is then automatically deployed via github pages.
 
 The shiny app can be deployed by running 
 `rsconnect::deployApp("shinyapp/DataExplorer")`, or from within RStudio by
 opening `shinyapp/DataExplorer/app.R` and clicking the 'publish' button.
+
+
+# TODO
+
+TODO.md has a list of planned changes. 
\ No newline at end of file
diff --git a/TODO.md b/TODO.md
new file mode 100644
index 0000000..f5baaf6
--- /dev/null
+++ b/TODO.md
@@ -0,0 +1,28 @@
+
+TODO
+
+This is a place for planned or desirable technical changes.
+Higher-level plans are discussed on
+the [Unjournal coda.io project management website](https://coda.io/d/Project-Management-UJ_dOyXJoZ6imx/Projects_subw9#Projects_tuA9I/r30&view=full).
+
+[ ] Render the quarto docs remotely via github actions.
+
+# Evaluating publication predictions
+
+[ ] Find out how to check publication automatically (ish?)
+  - get DOI; match it against databases; fall back to title and authors?
+  - map publications to "tiers"
+[ ] Build functions to create predictions from pairs of indiv. predictions
+  - simple average
+  - weighted by interval width
+  - extremized average
+  - for CIs, a Bayesian CI combining both?
+  - see Hanea et al.
+    "Mathematically aggregating experts’ predictions of possible futures"
+  - see aggreCATS package for some examples, but don't use it, unmaintained
+[ ] Test predictions against reality:
+  - both for our "aggregate predictions", and for preds from individual raters
+  - MSE or MAE for mean/medians
+  - coverage for confidence intervals
+[ ] Make a .qmd document or Shiny app reporting results overall, or for
+  specific subcategories of evaluations.
\ No newline at end of file