Notices by Ben Zanin (gnomon@mastodon.social), page 5

@staticsafe did you jut try it and it worked?

Nah its like a promo code that rarely gets advertised but usually stays active.

You can also try the medium code 3025 too

@staticsafe

or y'all can always shoot from the hip and see if code 4201 is active on any given day, and get a super discounted 4 topping large

if any of y'all in Canada need this, Domino's Canada has a 50% off thing going on for all pizzas at menu price for online orders only, valid till July 21st

code is CAG5014 but you'll see the offer on dominos.ca

[don't @ me with opinions about their pizza, it doesn't matter and I don't care]

2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97 #primeday

Matt Thorson (of Matt Makes Games, Towerfall and Celeste dev) is nonbinary!! once again bigots aren't allowed to have nice things

[17618.814804] mmc0: new ultra high speed SDR104 SDHC card at address aaaaaaaaaaaaa

of course there is no neutral writing interface and even—especially?—the qwerty keyboard is a *kind* of language model, expecting particular intents and producing particular kinds of text. and I do want to see a larger variety of writing interfaces serving creative, expressive, accessibility-oriented needs. but the ultimate teleology of tech like smart compose seems to be a world where "language" doesn't exist (only its statistical properties), and that feels gross to me

predictive language models have two actual applications: (1) non-deterministic sampling in poetry composition in order to suggest unusual juxtapositions; (2) creating text that resembles but is not language, in order to fill spaces that require or reward language-shaped things (earnings reports, blog spam, product reviews). smart compose is not doing (1), so its existence implies that my e-mail communication is among the language varieties described in (2), which it isn't

(a) a predictive language model by definition can only have output whose statistical properties regress toward the mean—that's the purpose of a language model in the first place, to determine how statistically likely a sequence of words is. (b) a language model is based on *text*, i.e., language ripped from context—so the output of a predictive language model (by definition) can't address shared emergent contexts between interlocutors

I really need to figure out how to articulate why "smart compose" in gmail fills me with visceral literally nauseating disgust

I resent that my computer, the most complex and powerful machine I'll likely ever use, is turned on by some pissant little five-cent plastic button like it's a shitty VCR from the 90's. I want a massive ornate copper knife switch, as long as my forearm, with the lower position hand-lettered "HALT" and the engaged position labelled "COMPUTE." When I go to use my computer, I want it to know I MEAN it.

Okay, updates.

A) None of the striking workers appear to have called on consumers to boycott Amazon during the strike.* So proceed in the way you believe best supports the striking workers.

B) sunglasses

C) The strike, currently, involves workers nationwide in Germany** and in Shakopee, MN, USA***. Other US Amazon workers do not appear to be striking.

D) The US strike is focused on free one-day shipping*, so--

* https://www.theverge.com/2019/7/14/20692187/amazon-strike-protest-workers-warehouse-employees-conditions-prime-day-2019
** https://www.reuters.com/article/us-amazon-germany-strike/amazon-workers-in-germany-to-strike-over-pay-verdi-union-says-idUSKCN1U90NZ
*** https://lifehacker.com/how-to-support-the-amazon-prime-day-strike-1836309627

Spoony boiz:

Also, here is a photo from last week of Nocturne gazing longingly at the tomato plant.

(He would later go on to not only chomp a leaf, but ingest a mouthful or two. And then he spent the next day feeling gross and hiding under the couch. I'd like to say he learned his lesson, but he did not—so now we have new rules about him being on the balcony.)

This Cosmo boy is all legs

look at this #cat