Jeremy Felt

A weekly note

Last week’s note was written in a rush and I barely proofread and oh my goodness the errors.

I think they’re fixed now. If not, oh well. It’s been a week!

I was making a coffee and reading others’ notes this afternoon and then remembered I still had mine to write. So a coffee and some clickity-clack it is.

Let’s see if I can finish before my coffee’s cold or gone.

Clickity-clack is what A says I’m doing when I go upstairs during the week to work. Michelle tells him that I’m solving problems, but one day I’ll explain to him the enormity of problems caused by all this clickity-clacking.

He sat on my lap at the keyboard on Friday and contributed a blurb of his own:

ghhhfhfhgyhuyuyr4r47uyr4r4utiio9][9o]po[‘]oop0
[o]’] cmjghity98757996

He then proceeded to hit a combination of characters that brought up an app selection screen that I can’t replicate for the life of me. I’ll be asking him for tech support before long.

A has been singing us his new song all week:

La la deee
La la deee
La la deee

Ah tooooo
Ah tooooo

It’s really great.

He’s also been requesting “old Donald farm” over and over again. He’ll sing the “old Donald had farm” part on his own, and then wait for us to do the rest. He’ll usually choose an owl or alligator or some inanimate object as the animal. Bananas squish, squash, squish if you’re wondering.

We’ve done a lot of Down Down, Baby lately. Our version is a collaborative mess of Nellie’s Country Grammer and the scene from Big.

I’m not sure if he’s actually “counting” yet, but he’s definitely saying 1, 2, 3, 4, 5, 6, 7, 8 while pointing at things in a set, so….. It’s wild to see a brain just go go go.

We saw the first daffodil of the year in Dry Fork last Sunday and the first daffodil of our front yard’s year today. All quite a bit later than last year. It’s nice to see things popping.

It’s still getting down into the 30s, but our days are sticking closer and closer to the 50s. Yard work time!

We had an electrician over to spec out the replacement of 13 ungrounded 2-prong outlet pairs on our main floor. I also walked through some seemingly strange decisions made by previous owners.

Our sewage ejector (exciting!) frequently trips the GFCI breaker it’s on, which it shares (????) with some indoor wiring in the bathroom. It’s very unnerving when you’re taking a shower in the basement and the lights go out because something connected to the water you’re standing in had an electrical fault of some kind.

I, an unqualified amateur, opened the “dangerous, there could be sewer gas” lid to the ejector last week and there was just a mess of water surrounding the pump above the tank and none of it seems right.

Anyhow. We want modern outlets on our main floor, that’s the real deal. But we’ll also get that wired on its own circuit and clean up a few other minor bits.

The estimate came in at about half of what we expected, which a very nice feeling when you spend the week preparing yourself for it to be double.

Next up is a plumber. In addition to the sewage ejector tripping the breaker, there’s an airlock in the piping that makes things shake in a way that makes you shudder every time it happens—several times a day.

A dime spent on the sewage ejector now is worth so, so, so, so, so much.

Finally working through some of this is very nice because I can start to visualize a not-so-distant future in which we’re rid of the gas-powered boiler, water heater, and range.

There’s a massive amount of visible piping in the basement used to route natural gas and boiler water that could be removed, making things much less pipey.

It would be interesting to be part of a large (large) language model poisoning network that works as a community to write nonsense targetted at over-eager language models.

When I first thought about that—obviously not the first to do so—I figured it would be relatively useless because you would also be “poisoning” any human readers with disinformation. For some reason I didn’t even think about white on white text, which a few folks have already used to inject prompts into Bing.

So now it’d be great to figure out the pattern for “invisible” text that is force-fed to LLM harvesters, but easily ignored by human visitors.

Pretty sure I just coined “large language model harvester”, you’re welcome.

And my coffee’s gone, time to make another and go outside. 🌻

Leave a Reply

Your email address will not be published. Required fields are marked *

The only requirement for your mention to be recognized is a link to this post in your post's content. You can update or delete your post and then re-submit the URL in the form to update or remove your response from this page.

Learn more about Webmentions.