The Amanuensis

  1. Trained on my library of articles on Instapaper,  1 the Amamnuensis will select articles from my Twitter feed and my Feedly feed. 2 These are in turn saved to Instapaper. 3
  2. Depending on level of confidence, some are marked as favourites (and so become public on Instapaper). 4
  3. Even more assortative links are shared on Twitter.
  4. The Amanuensis highlights interesting passages.
  5. The Amanuensis is trained on the books I’ve read.
  6. The Amanuensis is furthermore trained on books I want to read but have lacked the time or inclination. 5
  7. Based on its learning, the Amanuensis looks for new sources of reading upon which to train itself.
  8. The Amanuensis annotates and edits my texts.
  9. The Amanuensis writes my texts.


  1. 1Over 15,000 at the moment.
  2. 2Blog entry. The Amanuensis is here judging more deeply than I would. Or differently. While I choose articles based on whether the headline piques an interest educated by reading full articles, the Amanuensis first reads the article, then decides whether it’s suited. In some ways, it would be more “accurate” to let it decide based only on the headline and introduction. But that kind of accuracy is not necessarily what we want here.
  3. 3The Amanuensis will have a separate database, otherwise it will not know how to weight the links.
  4. 4Blog entry II. How does the Amanuensis choose important articles? Does it simply look out for certain keywords in common with other articles I’ve read? Most algorithms try to find what human users have liked. This sounds unimaginative but it manages to capture the free-wheeling nature of personal taste.
  5. 5Although still dependent, the Amanuensis is here beginning to “surpass” me.