Let's be Fwends #153:

Self-Defeating Thoughts, Screen Inferiority and Hooks

"But maybe destiny is always right now, right here, right this very instant, maybe."

~ Walter M Millar, A Canticle for Leibowitz

Hi, and welcome to Let's be Fwends number 153. This week, I have a great tip for you how to start productive mornings. We're also playing around with Large Language Models, learning about a technique to disarm unproductive thoughts, and get political towards the end.

As always, enjoy, and if you did, pass the word to someone who might enjoy Let's be Fwends as much as you do.

Having Action-Packed Mornings

No, not what you think (whatever it is you're thinking. You're wrong. I'm sure.)

But while we're talking: How are your mornings? More specific: How are your mornings at work? Here's a common scenario for me: I know I've got things to do, so I get up early, make a cup of coffee, and sit in front of my computer. Staring at the screen. Knowing that I should do stuff. Often even knowing what that stuff is. But still finding it extremely hard to get going. And starting with a shitty first draft does not sound very enticing.

Instead of wasting my otherwise perfectly good morning, I place some hooks in my work. Tiny little protrusions in the form of Todos that allow me to attach my attention to the work-to-be-done. I place them there the week or day before, for me to be discovered in the morning, when I'm tired and uninspired. A couple of things to do, well-thought-out by past me, for current me.

So, instead of going into the mentally demanding mode of shaping work, I start out with the mentally very easy work of ticking off boxes. And quite soon, I'm all warmed up and into shaping mode.

Asking an LLM to Write "Better Code"

Max Woolf came up with an interesting experiment.

He gave an LLM a coding challenge and prompted it into more and more extreme optimisations by simply asking it to "write better code".

After iteration 4, Claude (the LLM tasked with writing the code) finally moved into Product Management, claiming that the new version came

“with cutting-edge optimisations and enterprise-level features.”

But snark aside, Max saw an hundred-fold improvement in code execution time throughout the experiment (not that you could say the same about code legibility).

Who's Using LLMs and What For?

Speaking of Claude, Anthropic - the company making the model - recently released a paper analysing the conversations people have with Claude and deriving a couple of interesting facts from them.

Here's some interesting highlights:

1. There's a significant engineering bubble in regards to AI. Roughly a third of all usage revolves around software and similar technical tasks.
2. AI-usage does not adequately reflect occupational data. Just five industries (out of 22 tracked) drive AI-usage numbers.
3. Humans want to stay in the loop, but human-less automation is not far behind (57% vs. 43%)

(Source: The Anthropic Economic Index)

Don't be a Triceratops

Instead, be a Do-Ceratops. I sincerely apologise for this joke. I stole it from a cartoon.

What I'm trying to say: If you're prone to have self-defeating thoughts, Joe Oliver and Kristy Potter have some solid advice for you:

"(E)valuate thoughts not based on their truth, but on their workability. This means asking yourself: Does holding on to this thought help me live the life I want? Does it move me toward what is important to me? If the answer to these questions is ‘no’, it might be worth letting the thought go, even if it feels true. For example, the thought ‘I’m not good enough’ might feel true, especially if it echoes how you perceive your past experiences. But if this thought leads to self-doubt and avoidance (not engaging with a new hobby, skill, friendship, etc), it’s not workable. That is, it is not helping you live the life you want, nor is it moving you toward what is important to you. Instead of debating how true or untrue the thought is, you can recognise that it is not serving you – and decide to respond to it in a different, more skilful way.

I love the idea of not arguing with the thought, but instead considering whether it is useful, and if not, discarding it.

What is Screen Inferiority?

Text comprehension is lower on digital devices than with books. People reading textbooks on a digital screen find it more difficult to recall the content later, or explain key concepts beyond a very shallow level. That's a testable and repeatable result.

The interesting thing is: Nobody knows why.

Personally, I'm not surprised by this "Screen Inferiority". I find printed text easier to read - not the actual letters, but the text as a whole. When I open a book, or pick up a couple of printed out pages from a scientific paper, it feels as if I'm reading the whole text. I get lots of context. Where is the current passage located in the overall text? How many pages am I in, how many pages to go? What is around the current passage? Graphs, a new chapter, a summary, more text ...

In contrast, reading on a digital device often feels like reading with a magnifying glass. You see the thing you're currently working on really, really well, but struggle to understand the whole thing. You can't "zoom out", something that you do automatically when reading a paper.

I don't know if that is a contributing factor to Screen Inferiority, but it is something that makes reading on a digital device more challenging (but of course note-taking is so much easier).

We Are Heading For a New World

I find certain things remarkable. Trump and Musk in the USA, the AfD in Germany, the FPÖ in Austria, Le Pen in France, Orban in Hungary are dismantling democracy. And this is not hyperbole or an opinion. They say so themselves (even if many people choose to not believe them). They might argue that they will replace it with its stupid, ugly little brother Ochlocracy, but that doesn't change a thing.

What is astonishing that so many people (and the media) are still acting as if these people are still playing by the rules, while they explicitly say they are not.

So, I obviously was very thankful when I got Mike Brock' latest 'Notes From the Circus' newsletter, How to Stop a Coup.

The first thing is to acknowledge that it is a coup. And that lesson is applicable not only to the USA (which Brock is writing about), but also to Germany, Austria, Hungary, France, Italy.

This is not a test. This is real.

Experiencing Violence Creates Generational Trauma

Epigenetics is a strange thing. I hope I'm getting this (halfway) right: It basically says that outside factors change the way genes are expressed without changing the genome itself:

In response to stress or other events, our cells can add small chemical flags to genes that may quiet them down or alter their behaviour.

In a large study involving Syrian refugees (some of which survived the Hama massacre), researchers found that epigenetic changes where passed on to further generations:

In the grandchildren of Hama survivors, the researchers discovered 14 areas in the genome that had been modified in response to the violence their grandmothers experienced. These 14 modifications demonstrate that stress-induced epigenetic changes may indeed appear in future generations, just as they can in animals.

If your family experienced violence, excessive stress or any other trauma, it can affect generations to come. Not on a social level, but on a biological level as well.

Let's Be Fwends from the Past

Five years ago, nothing happened on Let's be Fwends, because it was on a hiatus for nearly half a year. What a shame. One year ago, there was a short but sweet discussion about metrics, proxies and why you really need to be careful about what you track.

That's it from this edition of Let's be Fwends. Take care.