Recently, there were two related threads on Hacker News: reMarkable 2.0 and note taking in 2021. In both mega-threads, much discussion centered around the nature of a Write-Ahead Log (WAL), without concretely referring to notes capture as such.

From a 10,000 meter overview, each of us are beset by information all day long; some of it we need to (or should) retain. "Notes" to some people mean the initial WAL, where we dump everything before routing that info to our calendar, general knowledge, or works-in-progress, etc. While to others, "Notes" mean our (hopefully permanent) knowledge base.

The WAL and the permanent data stores have completely different characteristics, which should drive what we use for them.

Write-Ahead Logs

The WAL is a record of our first contact with data. Before that data can become information, it needs to be reviewed and we need to think on it, so the WAL is also a scratchpad where we play with data trying to make sense of it. Then we can route the information to the best permanent data store.

The WAL should be cheap across all dimensions, ie. it should be cheap to buy, easy to carry, fast to use/read/write, etc. For me, this rules out reMarkables or other readers; as a pen and 80-page A5 notebook lets me quickly sketch out ideas without being overly careful about either the drawing or cups of coffee. Lacking that, markdown text goes to my wiki via my phone.

Permanent Data Stores

Typical (and perhaps "must-have") data stores are the likes of:

  1. Calendar -- punt data into the future
  2. Documents -- archive for reference
  3. Work in Progress -- add to your papers, research, whatever you work on

Since these absolutely cannot suffer catastrophic data loss, the typical risk analysis applies, ie. we need to guarantee their Integrity, Availability, and Confidentiality. For this, private git repositories work great since they are easy to back-up, and any file format can be dumped into them.1

That leaves the question of guaranteeing Integrity... and unfortunately in that regard, our biggest enemy is us. And our best defense is using a WAL to gateway and prove out data before we commit it to the permanent data stores.


The reviewing and rehashing of data into workable information could perhaps happen someplace else. For me, using paper is the easiest. However, that may just be an artifact of habit, and perhaps there are better ways of gatewaying Memory and her progeny, the Muses.

For published papers, a dedicated PDF reader seems already best... perhaps not so far off is a better WAL user-interface.