Posts in: tech

That feeling you get when something a long time coming finally does come out

I have always admired prolific writers like Matthew Yglesias and Scott Alexander — both now on Substack, and not by accident — for their ability to produce tens of thousands of words daily, My admiration being tampered somewhat by ChatGPT and other LLMs, which are about as intellectually and factually rigorous as Alexander, and slightly less so than Yglesias; some sacrifices do have to be made in the name of productivity. on top of the random bite-sized thoughts posted on social media. There are only so many words I can read and write in a day, and for the better part of the last year, my language IO has been preoccupied by helping clean, analyze, interpret, and write up the results of a single clinical trial, which are now finally out in The Lancet Neurology. Yes, my highest impact factor paper to date is in a neurology journal. Go figure.

The paper is about our clinical trial which used the body’s own immune system to treat autoimmune disease — and a particular one at that, myasthenia gravis — via technology that up until now has only been used against cancer (CAR T cells). It has made a decent impact since it came out less than two days ago. It got a write-up in The Economist, for one. Endpoints News as well. Evaluate Vantage got the best quote — it is at the very end of the article. And there is a whole bunch of press releases: from National Institutes of Health, University of North Carolina, Oregon Health and Sciences University, and of course Cartesian Therapeutics.

What went on yesterday reminded me that Twitter is not going anywhere any time soon: all of the above releases were to be found only there, not on a Mastodon instance, the journal’s own media metrics do not — and can not, at least not easily — trawl the Fediverse for hits, and I can’t just type in “Descartes–08”, “myasthenia gravis CAR-T”, or “Cartesian” into a Mastodon search box and get anything of relevance. One could, of course, argue that you wouldn’t get anything of relevance on Twitter either, most of the discussion consisting of people who have barely read the tweet, let alone the article. And one would be correct. And while most of the non-Web3/crypto tech world has moved out, it looks like people in most other fields, from medicine to biotechnology to the NBA commentariat, are maintaining substantial Twitter presence.

This will, of course, have no impact on my commitment to staying out of the conversation to the extent possible while maintaining a semi-regular schedule of 500-character posts, which may now, IO bandwidth having opened up, become a tiny bit longer. Thank you for reading!


Let the record show that I am bullish on Vision Pro, in part because of the glowing reviews from the likes of Ben Thompson and Matthew Panzarino, but mostly because of the misguided negative comments which are delightfully reminiscent of 2007. It has potential!


ChatGPT took their jobs. Now they walk dogs and fix air conditioners.

Writes the Washington Post:

Fein was rehired by one of his clients, who wasn’t pleased with ChatGPT’s work. But it isn’t enough to sustain him and his family, who have a little over six months of financial runway before they run out of money.

Now, Fein has decided to pursue a job that AI can’t do, and he has enrolled in courses to become an HVAC technician. Next year, he plans to train to become a plumber.

“A trade is more future-proof,” he said.

CGP Gray’s Humans Need Not Apply video is eight years old, but has never been more relevant. Leaning towards a profession more rooted in the physical world is a good instinct to have, for now.


By the way, the article I just wrote about found its way to me via Artifact, which has become great for discovering interesting content — thank you, clickbait-hunters — but subpar for reading it, even on the iPad. Anxiously awaiting a Reader mode there.


Having forgotten my work laptop at home — why oh why is every backpack I have black? — I now have to contend with two bizarro pieces of technology: a Dell laptop running Windows, and, after a long while, the Gmail web interface.

To my surprise, it is Gmail that is indisputably worse! Yikes, what a mess.


Project Tailwind — Google’s answers to DEVONthink, Obsidian/Roam, and ChatGPT rolled into one — has the potential to be the best thing that has happened to academic research in decades. Of course, coming from Google, I fully expect it to be gone and forgotten in a year or so.


Revisiting AOL Hell

The first article I ever saved to Instapaper — 11 years ago, three years after it was created and quite some time before it turned into my personal graveyard of text — was AOL Hell from the now defunct The Faster Times The link is to the Wikipedia entry which, strangely, says that as of February 2018 the url belonged to someone in Serbia. An odd coincidence.. The titular hell is the one AOL created for the employees producing content for its text mill. Writes Oliver Miller:

My “ideal” turn-around time to produce a column started at thirty-five minutes, then was gradually reduced to half an hour, then twenty-five minutes. Twenty-five minutes to research and write about a show I had never seen — and this twenty-five minute period included time for formatting the article in the AOL blogging system, and choosing and editing a photograph for the article. Errors were inevitably the result. But errors didn’t matter; or rather, they didn’t matter for my bosses.

This is, in fact, why AI will destroy the Web as we know it. And if you have any doubts about the outcome of Large Language Models being set loose on the internet, well, AOL wanted to do it even before it was trivial, destroying a few humans in the process:

The document reveals the same attitude that the bosses at the old Ford Motors factory had, when the assembly line was first introduced. Every week or so, the assembly-line was sped up; incrementally, barely noticeably, but the increase had a staggering, cumulative effect, and soon, those workers who couldn’t keep up found themselves standing by the wayside. If AOL could find a good way for machines to write about Lady Gaga, they would almost certainly fire the writers who remain.

They now have the machines, folks… They have the machines.

Mr. Miller can now be found on Medium, writing poetry.


A beautifully designed essay about an ugly entity: dark patterns. I’ve never heard of The Pudding before, but it seems like they do good work.


I just learned about Bike — from Brett Terpstra, who is back blogging and we are all better for it — and outlining will never be the same. Just look at the way it does text formatting, typing affinity in particular, and tell me you don’t want it in all your WYSIWYG apps (looking at you, Word).


Apps and/or services I have tried and dropped so far this year:

The one that stuck:

Yet again, Microsoft is eating everyone’s lunch. Back to the 1990s it is.