Whenever a system attempts to predict what a user of the system will want or do next or means, the failure case generates the most noticeable immediate problems. The author(s) will notice them and attempt to fix those problems.

The success case may or may not generate a long-term problem.

Systems in which this applies: autocorrect. Grammar checkers. LLMs. CPU branch predictors. There are others.

-30-

An extract from George O. Smith’s short story QRM INTERPLANETARY (1942), the first of his works about a space station set up to relay radio messages between Earth, Mars and Venus:

On Monday, Francis Burbank sent around a communique removing the option of free messages for the personnel. On Tuesday, he remanded the years-long custom of permitting the supply ships to carry, free, packages from friends at home. On Wednesday, Burbank decided that there should be a curfew on the one and only beer emporium. “Curfew” was a revision made after he found that complete curtailing of all alcoholic beverages might easily lead to a more moral problem; there being little enough to do with one’s spare time. On Thursday, he set up a stiff-necked staff of censors for the moving picture house. On Friday, he put a tax on cigarettes and candy. On Saturday, he installed time clocks in all the laboratories and professional offices, where previous to his coming, men had come for work a half-hour late and worked an hour overtime at night.

and about a month later it escalates to this conversation:

This is how I’m dealing with music playback in the house as of July 2025. My goals are reasonably high-quality music at a long-term reasonable price, while avoiding paid streaming services.

Since you’re interested in not breaking things, you should be running Debian Stable.

Debian makes several kinds of package changes available:

Hey academics!

Your papers, thesis, dissertation, project – whatever – is based on data, right? No data, nothing to talk about.

So every paper should have a data management section. It should be quite straightforward and easy to fill out, assuming that you were doing things well to begin with. (Hint: make it easy to fill out.)

• Data was stored prior to processing at this facility with the following access restrictions and encryption.

• Data was anonymized / is still live / was cleared by the IRB as not sensitive. Total data is X records.

• Data is archived at the university at this URL. Access is open / requires IRB clearance.

• Processing was done at these facilities. Intermediate results are archived / were securely deleted. Programs used were X (version), Y (version)… with the following licenses.

• No new code or configs were written for this dissertation / source code, executable artifacts and configs are archived at the university at this URL

-30-

Start with an Android phone with sufficient storage capability and a decent DAC/headphone amplifier. Built-in is great; otherwise, find something that works and doesn’t suck up too much power.

Add IEMs of choice, for less than $50, or cans that are reasonably efficient. It stops being so portable if you have 800g cans and a 500g battery-powered amp to drive them.

Install F-Droid, and use that to install Vanilla Music and RootlessJamesDSP.

Configure Vanilla Music to point to your music data.

Use ADB to enable RootlessJamesDSP (it needs 2-3 android system permissions).

Download convolution equalization for your IEMs or cans. https://autoeq.app is an acceptable and very convenient source. If Oratory1990 covers your equipment ( https://old.reddit.com/r/oratory1990/wiki/index/list_of_presets ) then choose that, first.

Apply in RootlessJamesDSP.

Enjoy.

-30-

When the Let’s Encrypt certbot says it failed and it was probably a firewall issue – yes, it’s probably a firewall issue. On your side. On port 80.

-30-

Specifically, a wiki which uses a non-database backend. I like Dokuwiki for this, but anything which basically saves in text files is fine.

(Why? Because you don’t need to bring up the database and/or fix the web server to grep for the page you need right now.) …

New paper: LLMs don’t do formal reasoning.

Well, of course.

LLMs don’t do informal reasoning, either.

Humans are great at pattern recognition. We even recognize faces in clouds and tree trunks. We recognize wheelbarrows, bears and crabs and archers in the stars. We make things that have meaning, and we communicate through speech and text and art.

LLMs are great pattern-generators. They are extremely well-tuned to make patterns that look like they might have meaning. A human trying to communicate may be bad at it, but they have an underlying model of the world that they are referencing and updating. An LLM is not trying to communicate anything. An LLM has a model of language, not a model of the world.

The map is not the territory. All models are wrong, but some are useful.

The situations in which it is reasonable to use an LLM are exactly the situations in which it is reasonable to roll some dice and use that to read the table of random monster encounters; to pull a card from an Oblique Strategies deck; to twirl the knobs on the synth and see if you can get a cool sound. In years past, you could type in a good list of keywords to Google and hit the I’m Feeling Lucky button.

Attempts to use LLMs for more than this fail. Often they do not fail in such an obvious way that the results are immediately discarded, which is where most of the danger resides.

-30-

Many organizations claim to be following Agile in some way. The vast majority are wrong, and most of them should not be Agile anyway.

You will recall that the Agile Manifesto, in total, is this set of four guidelines and an interpretation:

• Individuals and interactions over processes and tools
• Working software over comprehensive documentation
• Customer collaboration over contract negotiation
• Responding to change over following a plan

That is, while there is value in the items on the right, we value the items on the left more.

That’s a recipe for making certain customers happy with your perpetual engagement. But:

About two weeks ago I decided to use parts from a suddenly unused desktop to upgrade the living room media server. The initial plan was to keep the disks, power supply and case of the media server, and use the motherboard, CPU, RAM and video card from the desktop.

Things did not go to plan, of course…

My RSI prevention strategy involves change: I use three different keyboards, four or five pointing devices, and move around to different places. On my main desktop, I mostly use a thumb trackball, with a mouse in second place and a drawing pad in distant third.

On Friday the trackball died in a peculiar way.

I have almost 30 years of professional experience as a sysadmin.

Failure is inevitable. Hardware will fail, software will be discovered to have flaws, reconfiguration will be mistaken.

The antidote is to have reliable recovery mechanisms: everything will eventually break or need to be changed, so in turn you need to be able to recover back to a stable position, so you can try again.

Recovery mechanisms are not all-purpose. I usually describe them in three phases:

I had thought that ‘pareidolia’ referred exclusively to the human tendency to see human faces everywhere – any time you see two dots or circles and a line below them, boom :) it’s a face. It can be triggered by rocks, clouds, shadows, uneven browning on a tortilla or a pancake.

It turns out, though, that pareidolia covers any kind of overactive pattern matching. Humans look for meaningful patterns all the time, without thinking about the search, and when something trips the right thresholds in our brain, we seize upon it. The result can be faces, voices, music or just a cloud that looks like a moose.

Here’s my thesis: the large language models (LLMs) and similar generative software processes that people currently refer to as “AI” – those are exceptionally well-tuned to produce patterns that will trigger threshold recognition in human brains. Not by accident – that’s the goal. They are pareidolia-circuit stimulators.

The problem is that LLM output is exactly as meaningful as any other kind of pareidolia. Every well-formed paragraph is a false-positive error.

And when people repeatedly mistake their overactive pattern recognition for reality, we call that a mental disorder: psychosis.

-30-

If you have some sort of data or power cable that you tend to plug and unplug frequently, consider putting a sacrificial extender cable in between. The cost of replacing a jack on a motherboard or other fixed device is pretty high; the cost of replacing a plug on a device is often similar to the cost of the device in the first place.

For a headphone connection, for example, you can get a 6” extender for a cost less than the shipping. USB extensions are a little trickier - you need to match a bunch of different characteristics, but pretty reasonably priced when available. Power plug extensions are cheap, but it may take a little while to find one short enough to be reasonable.

-30-