An ongoing pattern in the AI march:

- the hardware efficiency improves
- the software efficiency improves

So I continue to argue that energy consumption as a PRIMARY criticism of AI is going to be unpersuasive, as incentives drive down the energy costs with time, and all critique of the energy costs is just as applicable to conventional cloud computing, which we have already accepted into our social bargain

(and which underpins every single tech salary)

so instead:

<https:redeem-tomorrow.com/the-average-ai-criticism-has-gotten-lazy-and-thats-dangerous>

See <https://pandoc.org/typst-property-output> for more info about the new [#typst](https://fosstodon.org/tags/typst) property support introduced in [#pandoc](https://fosstodon.org/tags/pandoc) 3.2.

[@mhoye](https://mastodon.social/@mhoye) I'm all in on “lightweight alternatives to Elasticsearch / Solr”. Let's see if I starred something useful…
<https:github.com/valeriansaliou/sonic> (Rust)
<https:github.com/zincsearch/zincsearch> (Go)
<https:github.com/askorama/orama> (JS)
<https:github.com/CloudCannon/pagefind> (specifically for static sites)
<https:github.com/kbrsh/wade> (Rust, library like Lucene)

The Dutch names for some of the elements/molecules are charming:
- hydrogen: water-stuff
- carbon-dioxide: carbon-stuff
- oxygen: acid/sour-stuff
- nitrogen: suffocating/choking-stuff

[@blabberlicious](https://toot.community/@blabberlicious) [@baldur](https://toot.cafe/@baldur) Corollary:
Experienced Writers: "Why would you use AI for writing? It's good at code generation tho."

Experienced Devs: "AI for code is garbage. I use it to summarize text, tho."

Eine Google suche kostet im extremen bis zu 0.0003 kWh oder auch 1.08kJ. Eine ChatGPT-4 Anfrage in dem Bereich von 0.001-0.01 kWh(3.6-36 kJ)

Also ist das 3-36 mal mehr um eine Antwort zu bekommen, die mit Glück richtig ist und mit Sicherheit nicht vollständig.

und alle so "This is awesome!"

ich komme da nicht mehr mit. Echt nicht.

shit shit shit shit shit. Dealt with my first set of AI-generated-code plagiarism cases this semester.

Dear students: LLMs are terrible plagiarists and they do not give a damn (or have any capacity to care or avoid doing so) if they get you in deep doodoo by reproducing verbatim code from existing repositories on github or from web pages. And on assignments, it's quite likely to happen.

We didn't identify these cases by using an AI detector: They were identical to a source on the web!

🖥️ New post! Yesterday I imported all my YouTube subscriptions into Feedbin so that I can watch videos via RSS. To do this I needed to turn my 500+ subscriptions into an OPML file. In this post I detail how I did it using a combination of awk, curl, jaq, JavaScript, shell, scraper, and Python.

<https:www.wezm.net/v2/posts/2024/youtube-subscriptions-opml/>

[#YouTube](https://mastodon.decentralised.social/tags/YouTube) [#Invidious](https://mastodon.decentralised.social/tags/Invidious) [#NewPipe](https://mastodon.decentralised.social/tags/NewPipe) [#RSS](https://mastodon.decentralised.social/tags/RSS) [#Python](https://mastodon.decentralised.social/tags/Python)

✍️ New post! I realised after mentioning the title case web page earlier today that I had not written about the other projects on my projects domain, 7bit.org. This post remedies that covering the following:

- Dew Point Forecast ­— 7 day weather forecasts that include the dew point
- RSS Please — Generate RSS feeds from web pages
- Titlecase — Convert text to title case
- MacBinary — a MacBinary decoder

<https:www.wezm.net/v2/posts/2024/7bit-projects/>

[#weather](https://mastodon.decentralised.social/tags/weather) [#VintageMac](https://mastodon.decentralised.social/tags/VintageMac) [#rss](https://mastodon.decentralised.social/tags/rss) [#dewpoint](https://mastodon.decentralised.social/tags/dewpoint)

I just watched Perfect Days (2023). I haven't yet fully understood the
effect it's had on me. But I think it's one of the best films I've ever seen. <https://en.wikipedia.org/wiki/Perfect_Days>

only way to regain sanity

Spectacular timelapse capturing an entire night from sunset to sunrise over the ALMA Observatory on the Chajnantor Plateau in the Chilean Atacama Desert.

Video Credit: ESO/C. Malin (christophmalin.com)

Timeline cleanse.

[#birdphotography](https://fediscience.org/tags/birdphotography)

When I think about what has happened with camera technology over the past quarter-century, I can't get my head around how it's possible to shoot a chestnut-backed chickadee at twilight at 1/100th second with a handheld 800mm lens.

[#birdwatching](https://fediscience.org/tags/birdwatching) [#birdphotography](https://fediscience.org/tags/birdphotography)

This cutesy feature popover in Slack is easy to dismiss until you discover that “precise customization" is the ability to disable the fricken’ gradients that make the navigation area such an unreadable mess.

Settings > Themes > Custom Theme > Window gradient — and you have to do it for every Slack you're in because of course you do.

Golden hour is for rabbits too.

Bought some Italian butter cookies this morning. Asked for an assortment.

Clerk: Any allergies?

Me (being silly): Just cats.

Clerk nods and made a show of putting back one of the cookies.

School tuck shop (aka canteen, cafeteria) menu from 1978, at the [#Australian](https://mastodon.social/tags/Australian) [#school](https://mastodon.social/tags/school) I attended as a kid. Older Aussies might find this interesting.

“Of course I like to read nontechnical books, although I read very slowly. Here are some that I heartily recommend.”—Don Knuth

😳

<https:www-cs-faculty.stanford.edu/~knuth/retd.html>

Another great tool is skhd (<https://github.com/koekeishiya/skhd>), which is a hotkey daemon for Mac. It's especially great for controlling things like yabai. Here's my keybinds for moving focus to different spaces, and also moving applications between spaces.

I’m never really a fan of covers, but Royel Otis' cover of Linger by The Cranberries is incredible. Feels like going back to the 90s/00s. <https://www.youtube.com/watch?v=JGUVB19e13s>

For once this bushtit held still long enough for me to squeeze off a couple of frames.

[#birding](https://fediscience.org/tags/birding) [#birdphotography](https://fediscience.org/tags/birdphotography)

I upgraded the llm-gpt4all plugin to support running Llama 3 8B Instruct (thanks, Nomic AI):

llm install --upgrade llm-gpt4all  
llm -m Meta-Llama-3-8B-Instruct 'Write Python code to print 5 great names for a pet pelican'  

The quantized model is a 4.34GB download and should run on machine with 8GB RAM - works great on my M2 MacBook Pro

<https:github.com/simonw/llm-gpt4all/releases/tag/0.4>