Daniel Tubb

More revising. It’s feeling quite workman like. Which I love. What section am I working on? Decide. Where are my notes? What have I already written about this? What is duplicated text from various drafts? What ideas go together? Put them together in a drat. Cut, edit, reorder. Rewrite. Rewrite. I’m done. Rinse and repeat. Repeat.

Today, I did a section on Wendy Belcher’s brilliant book, Revising Your Journal Article in 12 Weeks (525 words), and then a paragraph on books on grammar (215). Plus, a bit of a critique (95 words). And a paragraph on what this book is not (100 words.) Doesn’t seem like a lot. But, but it’s good.

I think tomorrow I can go a lot faster.

EasyFind and BBedit are the tools for this.

I’ve been revising, editing, and removing duplication for the Reader’s Guide. A section basically summarizing some of the excellent work out there that is how-to advice for writers. Of course, this book is no how-to book. But sometimes it’s important too know what you are not.

Also, spent a bit of time working on a script that integrates Bookends to CiteProc with Marked. Works wonderfully. You can write with Bookends in Markdown, and get back CiteProc, or move it to Word or Mellel. Took some fiddling, but I’m very happy. It’s perfectly formatted, Chicago Manual of Style 17th edition for books.

I’m not good at this daily writing. But, yesterday, while I didn’t do a blog post, I did do a lot of coding in the morning. Today, I did as well. A Reader’s Guide section, on this book as not self help, which of courses summarizes lots of the books I’ve read over the years.

But, I’m really excited of the process. Bring together text, code it, cut repetition, organize it, write it. It’s working.

Today, I did 1800 good words. Plus, coded 30,000 words down to 20,000. This week I can just go through and cut, edit, delete, them into shape.

Progress. At last.

Took Sunday off. So much for posting every day.

In any case, today I sat down with a yellow legal pad, a newfangled old-fashioned typewriter, and thought about the form of this book. What exactly do I mean by a castnet, an atarraya?

The form? A book of essays, that re woven together. Each is about 20 pages long. Give or take. No more or less than 17 and 25.

Doing that, forcing my self to slow down, was enough to let us sketch out the “spokes” or “radial threads” of a castnet. There are eight. These are probably going to feel very much like chapters. In the centre, there’s the Reader’s Guide. Outside that, on the first rung, we have the crisis on the linoleum, that will fead into the 8 chapters. It will be an interlude between each chapter. Then we have the threads: First and Second Fieldwork, Time and Money, Tools, Rules, Fichero, Bricolage, Makeshift, and Ambulatory Anthropology. Each thread has related scenes and arguments. Acknowledgements will come at the end.

Perhaps length and argument something like:

• Reader’s Guide (atarraya as social conversation) (2,000)
• Crisis on the Floor (writing is the hardest part of the deal) (7,000)
• 1. First and Second Fieldwork (cane toad) (7,000)
• 1. Time and Money (room of one’s own) (7,000)
• 1. Tools (panning into various tools for writing) (7,000)
• 1. Rules (mad rules (e.g. 789 serialized and don’t break the chain ) for writing) (7,000)
• 1. Fichero (mining into zettklekasten) (7,000)
• 1. Bricolage (workshop and improvise) (7,000)
• 1. Makeshift (workshop, and good enough mastery of non-mastery) (7,000)
• 1. Ambulatory Anthropology (embodied writing) (7,000)
• Acknowledgements (2,000)

Say, (2000×2)+7,000+(7,000×8 chapters) + 2,000 notes + 2,000 citations = 72,000 words.

I spent the first 2 hours doing that, and then 2 hours putting it into Tinderbox.

Structure is the hardest part of the deal. I was inspired with writers method of working long hand, with a pencil, and retyping.

Wrote a brief piece for the newsletter on writing without agency. On de-centring the idea that writing is a product of rational thought, rather than the thoughts themselves. Thus, the words have a certain agency. Probably, this is one of the flaws of thinking that AI is going to be a replacement for writing. Not the best piece, but a first draft certainly of an idea I had mulling around my head for years. So many times, I’ve gotten into my atarraya archive to do one thing, and instead I’ve done something else complete. Ideas have sparked off other ideas, that were put together by accident. Where is the agency? Order and control come out of the work.

Late post. All of it, the newsletter and this post, took 25 minutes.

789 is a good number.

Also ran 13.5 km today, and got Fichero building on Windows, Linux, and launching (then crashing) on iOS and Android.

Good day.

Did a lot of coding late last night: tagging and categorizing and putting into categories, things I have written on bricolage.

Then, I found in the corner of a window on the bottom of my screen, a piece on fieldwork and participant observation. I realized it would fit well with the section I wrote a while ago on makeshift. So, I read that piece quickly, made a couple of minor edits, connected it to the piece on makeshift, and sent it out on 789 Serialized.

I’m still concerned about focus and structure and the whole book. But, that’s not a problem that can be solved today. And it’s probably not best solved at all in the abstract. Rather, it’ll be best solved as I get pieces tightened and written. And as I keep the momentum going.

A book is written not in one day.

I’ve got other things to do.

Today was a slow day to get going. E-mail. Office demands. Tomorrow, no e-mail before noon.

But I did get to coding the book for a few hours in the evening, and then edited a section on Ferdinand Cheval’s Ideal Process. I’m not sure where it will go in the end. But it feels like it’s connected to the workshops. So I put it there.

The method worked. I had written pieces a few months ago, but I went into the archive. I found all related text to Ferdinand Cheval’s groto in there, which he made by hand over 30 years.

I’ve not watched the movie yet. I might do that tonight. ((L’Incroyable Histoire du facteur Cheval)[https://www.imdb.com/fr/title/tt7248884/], 2019).

How does it fit into the big picture? I’m not sure. Does it matter? 1473 words. 684 more than I needed to.

Just posted my second 789 Serialized post. I’m not super happy with how I’ve set it up on Substack. But, no matter. I can move over in the future to something else. The part I’m not sure on, is a platform for multiple newsletters. But, anyway.

The piece is on the greek concept of practical knowledge and mētis. I sat down to code some text, and it’s the first file I opened. Rather than code it, I edited it, and made it much better.

It’s nice to something off, as I fell behind the last two days. Too many other writing tasks.

I also spent a bit of time updating my logging for this project. It’s a markdown file, that a shell script updates when I add a file to the filed folder. I could set that up to work with folder actions, so when I make a change in the finder, the script runs, and the website is updated.

In the meantime, I’ll post it manually.

Also, as I filed this piece, I realize it fit right after the section on the attarya and fishing in Mompox, which is the centre of the book. An accident, but I think a nice serendipitous place to put it.

David Graeber once tweeted, at least I remember reading it but can’t find it, that he used to have two projects to work on. The A project, the thing he had to do, and the B project, the thing he used to procrastinate on the A project with.

My B project seems to be what I’m working on this morning, rather than the 789 Serialized. An article with a colleague about artificial intelligence and the Dawn of MIDI. We’re calling it the Dawn of AI. Its argument is at once a critique of how artificial intelligence is being used by students, and at the same time, it is a call to consider AI as offering the possibility of at once enabling new kinds of research that perhaps didn’t really need AI in the first place.

Re-reading the article, it was clear it had to do one or two things differently. It needs to be concise, and it needs a home.

Reviewing the draft we had, there were just too many examples. We had 5, but space for 1 or 2. Thinking about it, I can see two examples that have a synergy and fit together: vibe coding, on the one hand, and then using AI to read archives.

On a home. Today, before revising, we settled on an audience. A few journals come to mind. In both cases, the length is Chicago Manual of Style and 8,000 words seems appropriate.

So, 2 points, 8,000 words, CMOS, and it seems more doable.

This morning, I also drafted some notes on rules for writing.

Also, I missed this and the newsletter for yesterday. Hah. So much for Mad Rules for Writers.

I ran this morning, 7.3 km. I changed the route and headed north from the village. It has a hill! I ran downhill for 2 km, then uphill for 2km. It felt good to be out. I came back, showered, shaved(!) and sat down at the computer.

I edited a prologue section. It will become today’s 789 Serialized. It actually became three or four. I need to decide. Is 789 words the minimum or the maximum. I think it feels too little, this morning.

On the run, I thought about arbitrary rules for writing. I have had many over the years. I realized, yesterday. The point is I set them. They work for me, for a time. The particular don’t matters.

Craig Mod talks about his rules. A portrait on his walks by 10 a.m. A pop-up newsletter post a day. Mad rules. But stick to them.

Mad rules for me, are not for anyone else. What is 789 Serialized but some mad rules for me to write every day for a month? At least 789 words a day. But, maybe more.

On the run, I took notes for my self-evaluation. It was due a month ago, one of those academic chores I resent. I need to get it in. But I wonder if I think of these chores the wrong way. Why not see them as something to share? To publish as a newsletter. Online, publicly. Online, privately. I’m not sure. But think of an audience as more than an overworked administrator who may or may not read it.

For this book writing diary newsletter, my thoughts turn to the problem of coding huge quantities of text—field notes, morning pages, book drafts, articles, essays, old chapters, pieces of text, blog posts, articles. All of this is my archive. My fichero, as I’ve come to think of it. (Fichero from the Colombian for a box of index cards). It is now about 500,000 words coded, and another 3,000,000 words uncoded. How did I write that much? Why did I write that much.

One, is that I’ve accumulated text files of repeated text. Different versions of the same thing. Different drafts. Cruft that has duplicated. How to deal? Cut and don’t make more.

On cutting, six months ago, I wrote a script to delete second and subsequence instances of the same words, sentences, or paragraphs. Dejatext. I use it all the time.

On not making any more duplicated text, the issue is how to stop creating duplicate words. I have another script I use to code text. It takes text I’ve coded, and splits it into new text files, and brings together codes. It works well. I run it, and organize my notes. It’s good. But, one of the mistakes I’ve made over the years is letting repeated text pile up. Drafts and drafts and then I forget which is the most recent.

One issue. A reticence to delete. I leave words in the archive, even as I use them. Why? Why not take the text, code it, move it from the archive, synthesize it, publish it, and move it along.

How should this workflow function? What’s the actual process? Should I edit the original text file, or the coded version? It seems most logical to edit the coded text—then discard the original. That way, I can iterate toward order over time.

The key question is whether structur.py supports that approach. I may need to revise it.

How would this work? An inbox of notebooks and other sources; a to-code folder with raw uncoded text files; a coded folder with text organized by code; a codes.txt file listing possible codes; and a synthesis folder where ideas are developed. This way, I’d not have duplicating text.

Right now, I’m not sure if structur.py can append new material to existing code files. That’s what it needs to do: allow new material to be added to existing coded text. The coded folder should be able to receive more material over time.

Another issue: how does it handle text that’s already been processed? One possibility is that the system always regenerates empty coded files from codes.txt. If a code is moved or deleted, it’s recreated empty—ready to receive new codes text. That might make sense.

In that model, codes.txt acts as the master list. The coded folder always mirrors those codes. Structur recreates empty code files if needed, letting me keep feeding in new material. This would support then an iterative process of coding and synthesis.

A project for after finishing my 789 Serialized post for today.

Update: 8:30 PM: I fixed structur.py, squashed a bunch of bugs, posted it to GitHub. I also added unit testing to dejatext. Structur now works more reliably, and I deleted some duplicates, and pulled out some malformed text. Long day. But, good changes.