Log-level thoughts (HTML , how to share slop , AI is like a really big number)

Benefits of HTML:

• It's a bookmark

Your work is never done. So a nice, summary-level html document is a way to help future you (or someone else!) pick up where you left off.

• It allows the LLM to be more expressive

A diagram can be worth 1,000 words. Allowing the LLM to make you diagrams means more diagrams, less words. You have a big visual cortex, html helps you use it.

Anti-benefits of HTML:

• it's slower

See chart above. I also have a skill /sync that instructs the LLM to NOT generate HTML because ... I'm right there, on the other side of the keyboard, waiting for a response.

• it's weirder

This is also arguably a benefit. When the LLM generates slick HTML for you, it is very obvious that there is not a person on the other end of the line (unlike chat, which very much implies a person-like presence). But it is weird. Especially at first. It feels indulgent and bizarre.

Introducing...the Slog!

My slog is a slop blog.

Working towards a world where:

• 1: I read a blog post, and it gives me an idea
• 2: I use an LLM to get that idea out of my head and into HTML
• 3: I publish it as a slog post
• 4: Someone else sees the post, pulls down the zip + then riffs on it and publishes it to their slog or equivalent system
• 5: We all become friends

How I'm thinking about AI on July 25th 2026

AI is like the number 1,000,000,000

I don't really know what a billion is.

I know that 1 Billion + 2 Billion = 3 Billion. I can watch a youtube video (or three) with some neat visualizations about how much land you could cover with a billion dollars in $100 bills or whatever. But I don't really it. The best I can do is remember that I truly don't get it, and train myself to use calculators + code.

AI is a really big deal. What kind of big deal? I have no idea. Best I can do is remember that I truly don't understand. I deeply, profoundly, have no idea how LLMs work and best I can do is think hard about context, blast radius, and what it means to be a person and live a good life and help people in 2026 (and beyond).

If I got to participate in a world-historical moment of madness/utopian possibility/topsy-turviness I wouldn't tell anyone...but there would be signs.

"Understand to participate" is the best idea of this moment

https://www.geoffreylitt.com/2026/07/02/understanding-is-the-new-bottleneck

Geoffrey Litt should rename his talk/post Understand to Participate (like Simon Willison said when he posted about his talk).

When the doctor explains some organ to you, it's not because the doctor wants your intellectual contributions to medicine. The doctor is explaining the organ to you because it's your organ, your body, and so you should be in the loop for tricky decisions that will involve tricky trade offs.

If we aren't using LLMs to include more people in our little software games, what are we even doing?

Another potential quippy phrase:

(Sometimes) The User Is the Human In The Loop

Sunk Cost Fallacy of Software Engineers is a big part of this moment in time

The people who are most exposed to LLMs right now are people who have status in the software world. By definition, a lot of these people are dealing with a massive sunk cost fallacy.

The world now works differently. It's not just about learning new things but about unlearning old things.