iiinova — where your digital self keeps evolving

jneen @jneen.ca · 15/07/2026

ok bsky devs i have a gripe about the video player. see that volume slider? see what's right behind it? guess what happens when you click the volume slider

A bluesky video player showing the picture-in-picture button overlapping the volume slider

130

jneen @jneen.ca · 14/07/2026

it really do be like that

github file display screenshot, showing two files with commit messages.

.claude (Initial import.)
.github (Funding update.)

030

jneen @jneen.ca · 02/07/2026

this is what being a musician is like. both of these companies are making bajillions.

Paypal history:
Bandcamp Ventures LLC +4.81 CAD
SoundCloud Global Limited & Co. KG -47.32 CAD

1111

jneen @jneen.ca · 28/05/2026

there's a reason we called out programming specifically in Automaphonica's AI policy: automaphoni.ca/ai-policy/ and the reason is that programming is a creative task. the "routine work" in programming is no different than the "routine work" in any other art.

Works created using Generative AI, Large Language Models or Diffusion Models are NOT accepted in:

* Composing or producing Automaphonica music
* Programming or producing video for Automaphonica
* Producing art for Automaphonica
* Hosting or promoting Automaphonica releases

1131

jneen @jneen.ca · 26/05/2026

Mayor Mamdani sits down with violence

120

jneen @jneen.ca · 24/05/2026

is signal's username feature real or is it like, an elaborate hoax

Your username couldn't be saved.
Check your connection and try again.
[OK]

000

jneen @jneen.ca · 23/05/2026

picture of RetroArch trying to open a GBC file:

[0: jneen@lavender ~ ] $
; /Applications/RetroArch.app/Contents/MacOS/RetroArch
Segmentation fault: 11 /Applications/RetroArch.app/Contents/MacOS/RetroArch

010

jneen @jneen.ca · 23/05/2026

what on *earth* is this supposed to do other than tell me no program can open files with your random ass extension. the app has no "open" feature. do people know how to write software?

Obtain your patch file and create your ROM

When you join a multiworld game, you will be asked to provide your config file to whoever is hosting. Once that is done, the host will provide you with either a link to download your patch file, or with a zip file containing everyone's patch files. Your patch file should have a .apladx extension.

[highlight]Put your patch file on your desktop or somewhere convenient, and double click it.[end highlight] This should automatically launch the client, and will also create your ROM in the same place as your patch file.

110

jneen @jneen.ca · 22/05/2026

Extremely good and concise summary of the statistical/methodological shenanigans involved in measuring the so-called "AI boost" Here's a couple of my favourites:

Timing Artificial Tasks

A widely cited study found that developers who used GitHub Copilot completed a task 55% faster than those who did not [Peng2023]. The task was implementing an HTTP server in JavaScript from scratch, in ninety minutes; the developers had no other obligations that day. Real software development involves navigating a large codebase you did not write, understanding a requirement described ambiguously in a ticket, coordinating with colleagues, and attending meetings. Speed on a greenfield toy task does not predict speed on any of that. A randomized controlled trial with experienced open-source developers found the opposite of what participants themselves predicted: giving them access to AI tools increased task completion time by 19% [Becker2025].

Comparing AI to Nothing

Studies that compare AI-assisted developers to a control group using nothing have chosen a baseline that does not exist in practice. Developers without LLM assistants use documentation, colleagues, and the time they would otherwise spend thinking through the problem themselves. The relevant question is whether LLM tools outperform the alternatives developers already have, and that comparison is rarely made [Peng2023]. Choosing a weak baseline makes any tool look good; it does not make the tool useful.

042

jneen @jneen.ca · 22/05/2026

i hate this fucking picture

Norman Rockwell Guy Standing Up meme, commonly seen on the worst fucking takes you've ever laid eyes on

140

jneen @jneen.ca · 21/05/2026

one of the best parody accounts to ever do it. i wish they were here. nitter.net/outingpredat...

Tweet, 16 Jun 2018 from "Goku" @outingpreditors

Slammming my fists onto my Microsoft Surface 2 with 64GB Memory because the Croc wikia refuses to tell me how old he is .

1 Reply, 115 Retweets, 212 Likes

010

jneen @jneen.ca · 20/05/2026

this the right one?

The Next Revolution In Industrial AI Has Arrived.

Because The Best Intelligence Isn’t Artificial. It’s Human Intelligence, Amplified.

SeeQ(A)I

200

jneen @jneen.ca · 13/05/2026

you don't need AI for this. we have grep

The developers I talked to found AI useful for some tasks. Several developers said that it was good for experimentation, allowing them to quickly prototype an idea or to implement something in a domain they’re unfamiliar with. One developer said it was a good information interface. Specifically, he said, [highlight]the AI helped him find where on the server a certain request is handled, summarize logs, or find documentation related to code changes.[/highlight]

110

jneen @jneen.ca · 13/05/2026

this is what i've been SAYING

“We've been told performance evaluations are tied to AI adoption,” the UX designer told me. “This has led to most of my teammates using it performatively, even if most of us implicitly know that the output is flawed. [highlight]The actual quality of output doesn't matter as much as our willingness to participate.”[/highlight]

271

jneen @jneen.ca · 08/05/2026

"pseudoscientific" is in the first sentence of the wikipedia page on polygraph tests btw

A polygraph, often incorrectly referred to as a lie detector test,[1][2][3] is a pseudoscientific[4][5][6] device or procedure that measures and records several physiological indicators such as blood pressure, pulse, respiration, and skin conductivity while a person is asked and answers a series of questions.[7] The belief underpinning the use of the polygraph is that deceptive answers will produce physiological responses that can be differentiated from those associated with non-deceptive answers; however, there are no specific physiological reactions associated with lying, making it difficult to identify factors that separate those who are lying from those who are telling the truth.[8]

140

jneen @jneen.ca · 03/05/2026

linux discourse on the tl today

tweet from @shaun_vids 28 Jan 2020: brigaded by the vile trolls again, simply for deliberately provoking them with lies and insults

030

jneen @jneen.ca · 29/04/2026

absolutely wild

VERSIONS:
0.0.3 April 28, 2026 (8.5 KB)
0.0.2 November 10, 2013 (7.5 KB)
0.0.1 November 10, 2013 (7.5 KB)

000

jneen @jneen.ca · 24/04/2026

that's it? that's all we get? "some pull requests may have been merged incorrectly"?

Update
We have resolved a regression present when using merge queue with either squash merges or rebases. If you use merge queue in this configuration, some pull requests may have been merged incorrectly between 2026-04-23 16:05-20:43 UTC.

120

jneen @jneen.ca · 23/04/2026

if i become a math teacher i'm using this tweet to explain injections and surjections

tweet by Kusi Constanza @AntifaConstanza:

There are more grains of sand on the beach than atoms in the universe

100

jneen @jneen.ca · 18/04/2026

job postings at home:

A model is only as good as its data, and beneath every breakthrough in AI, there's an orchestra of human expertise powering that data.

micro1 conducts that orchestra, finding the sharpest minds and forging their expertise into the datasets that shape how AI reasons, adapts, and integrates into the future of humanity.

000

jneen @jneen.ca · 17/04/2026

i know it's not the point, but what is this kind of journalism called

We have both ordered crab chowder as a starter. Two waiters arrive and with a dramatic flourish pour the soup over the shredded crab and deposit a shared plate of four mini rolls between us. The chowder is spectacularly good, both smooth and zingy. Both of us stick monastically to sparkling water.

140

jneen @jneen.ca · 15/04/2026

average computer user looking at a new tool

"this better not awaken anything in me" meme with the Dean from Community, but captioned "this better not make me learn anything..." in green monospace text

020

jneen @jneen.ca · 01/04/2026

And in this thread you seem to dismiss what is fairly clearly the only reasonable course of action - reverting the theft of the repository - as "not serious" with precisely zero justification.

Reddit comment by scheems (OP)

My position is that giving RubyGems.org to gem coop won't be happening. My position is that the suggestion that Ruby central fork RubyGems.org codebase to run it was not an intellectually serious proposal. And not a good idea.

010

jneen @jneen.ca · 26/03/2026

there is truly nothing new under the sun

If there is such a person as a program coder, and I have heard of *one*, then he is going to have to change his job — perhaps running The Last One on behalf of the analysts? The analyst/programmer should welcome this development because it means that he can spend more time on ideas and less on tedious coding and debugging. System designers [cut off]

000

jneen @jneen.ca · 26/03/2026

job adverts at home:

030

jneen @jneen.ca · 21/03/2026

mark levine, my dude, this is still the first chapter of the jazz piano book, you're still explaining what intervals are. why is this wild ass voicing your example 😭

piano music staff. chords: G7b9 Cmaj

chord 1: left hand: F B right hand: E G# B
chord 2: left hand: E A right hand: D G C

the top B and C are connected by a line annotated "minor 2nd"

330

jneen @jneen.ca · 20/03/2026

i know this point has been hammered into the ground by now but "shipping more often" is a terrible awful metric to optimize for. by far the fastest-deploying team i've ever worked on was the one that was constantly firefighting on account of our garbage codebase and infrastructure

anthropic-ai (Promoted)

What does your team's deploy frequency look like? Claude-using teams on Vercel are shipping 7.6x more often than those without it.

3224

jneen @jneen.ca · 16/03/2026

*please* correct me if i'm wrong, but i'd back-translate these as "i learned from his podcast that there's a high school student who identifies as a cat" / "in that case I identify as a bearded dude" ...which is just a worse joke? maybe i just don't understand japanese comedy

101

jneen @jneen.ca · 14/03/2026

In particular, Karapathy makes the fairly strong assertion that hallucination is not different than normal output, which is because the architecture (largely the same as ChatGPT and other LLMs) only "knows what sequences are statistically plausible".

What’s the deal with “hallucinations”? The model generates tokens by sampling from a probability distribution. It has no concept of truth, it only knows what sequences are statistically plausible given the training data. microgpt “hallucinating” a name like “karia” is the same phenomenon as ChatGPT confidently stating a false fact. Both are plausible-sounding completions that happen not to be real.

210

jneen @jneen.ca · 13/03/2026

alright pack it up, AI systems already obsoleted programming in the 80s

(Wikipedia screenshot)

**The Last One** is a computer program released in 1981 by the British company D.J. "AI" Systems.[1][2][3][4][5][6][7][8][9][excessive citations] Now obsolete, it took input from a user and generated an executable program in the BASIC computer language.

The name derived from the idea that The Last One was the last program that would ever need writing, as it could be used to generate all subsequent software.

The software was a program generator, as distinct from an actual programming language, as programs were generated by the user selecting options from menus that would form the basis of the generated code. This was done in a logical sequence that would eventually cause a program to be generated in BASIC. At any time, the user could elect to view a flow chart showing the current progress of the program's design.[10]

The program was available on various different computer platforms of the time.

(side card)
The Last One
Original author David James
Developer D.J. "AI" Systems
Initial release 1981; 45 years ago
Type Programming language

010

jneen @jneen.ca · 10/03/2026

so i'm lookin at the khronos docs site to use as a source for glsl builtins, and...

$class ExtendedRougeSyntaxHighlighter < (Asciidoctor::SyntaxHighlighter.for 'rouge') register_for 'rouge' # Insert rouge stylesheet from super # Then replace many 'github' theme colors for accessibility compliance # It would be better to use rouge's stylesheet factory, if it has one def docinfo location, doc, opts overrides = %(<style> /* Khronos overrides for Rouge 'github' theme for accessibility */ [highlight]/* Basically everything is overridden, but it is unclear how to add a new Rouge theme */[end highlight] /* Codelike overrides */ pre.rouge .cm, pre.rouge .cp, pre.rouge .c1, pre.rouge .cs, pre.rouge .c, pre.rouge .ch, pre.rouge .cd, pre.rouge .cpf, pre.rouge .gh, pre.rouge .bp { color: #5f5f5f; }$

020

jneen @jneen.ca · 09/03/2026

we have invented ELIZA, from the popular book "don't invent ELIZA"

[highlight]The chatbot he used was Chai’s default personality, named Eliza[end highlight]. It shares a name with the world’s first chatbot, ELIZA, a natural language processing computer program developed by Joseph Weizenbaum at MIT in 1964. ELIZA responded to humans primarily as a psychotherapist in the Rogerian approach, also known as “person-centered” therapy, where “unconditional positive regard” is practiced as a core tenet. The researchers working on ELIZA identified from the beginning that their chatbot posed an interesting problem for the humans talking to them. “ELIZA shows, if nothing else, how easy it is to create and maintain the illusion of understanding, hence perhaps of judgment deserving of credibility,” Weizenbaum wrote in his 1966 paper. “A certain danger lurks there.”

151

jneen @jneen.ca · 09/03/2026

nice! i've been thinkin about this one

120

jneen @jneen.ca · 08/03/2026

oh i also forgot about "pool", a slightly deeper version of "chilly" from the original colour set

000

jneen @jneen.ca · 08/03/2026

over the years i have added a few colours * "krasna-light" is a lighter version of the background colour * "dragon" is a nice deep green that i think goes well on krasna * "rouge" is Edward Loveall's pinkish red from the original Rouge site I think these round it out a bit, esp for terminal use.

100

jneen @jneen.ca · 05/03/2026

i might be cookin a little

$[website screenshot, with pleasant syntax-highlighting colours featuring a deep blue background] Rouge An elegant, extendable code highlighter written in pure Ruby. Available on GitHub Try some code! [Ruby] [main] class Greeter def initialize(name="World") @name = name end def say_hi puts "Hi #{@name}!" end end Result: [the same code as above but nicely highlighted] [Save this]$

140

jneen @jneen.ca · 03/03/2026

this "pretty convoluted" comment has haunted me for a decade. can you guess what they want instead

# bad - pretty convoluted
kind = case year
when 1850..1889 then 'Blues'
when 1890..1909 then 'Ragtime'
when 1910..1929 then 'New Orleans Jazz'
when 1930..1939 then 'Swing'
when 1940..1950 then 'Bebop'
else 'Jazz'
end

010

jneen @jneen.ca · 03/03/2026

imagine having a linter that someone thought about for more than 10 seconds

$highlighted Ruby code, indented very strangely ``` private_class_method def self.parse_cgi(str) pairs = URI.decode_www_form(str).map { |k, v| [k.to_sym, v] } Hash[pairs] end ```$

120

jneen @jneen.ca · 02/03/2026

Hi there, computer scientist here. Does the difference between "plausible" and "true" matter in labour law? Because systems like this are hyper-optimized for plausibility, not for truthfulness, and this is an important distinction. Ask leading AI researcher Andrej Karpathy if you don't believe me.

[bold]What’s the deal with “hallucinations”?[end bold] The model generates tokens by sampling from a probability distribution. It has no concept of truth, it only knows what sequences are statistically plausible given the training data. microgpt “hallucinating” a name like “karia” is the same phenomenon as ChatGPT confidently stating a false fact. Both are plausible-sounding completions that happen not to be real.

100

jneen @jneen.ca · 27/02/2026

dam i didn't think it was that bad up here yet

030

jneen @jneen.ca · 26/02/2026

and one of the worst things you can do is put us on PIPs or push us out because we're "not being productive". you have *no idea* what it is we go through every goddamn day. welcome to being trans. all our friends are on PIPs.

edited book cover with a confused-looking wide-eyed brontosaurus, titled "All my friends are [edit]on PIPs[end edit]" / Authors Avery Monsen and Jory John

120