If your world view consists of white cowboy hats and black cowboy hats you may want to check whether your white cowboy hats aren't just black cowboy hats that have been spray painted white.
Pinned post, about following from multiple addresses
Hi, when I follow someone I usually do this from multiple accounts, with different servers and platforms. This is partially as backup. It’s also because different software has different features. E.g. at the moment of writing I mainly post from my Friendica server, but it doesn’t yet support polls, or account migrations. So I also follow from my Pleroma server, which does notify me someone has migrated.
I also sometimes use platforms that support nomadic identity, that allows my channels on multiple servers to be synced. From software that doesn’t support that, a follow request looks like multiple separate requests from those channels.
Hi, this me from my Friendica instance. I also post from @GidiKroon@mastodon.social, @gidi@pleroma.gidikroon.eu as well as from Twitter, Instagram and Tumblr as @GidiKroon but I'm mainly using my own Friendica for now. Even though I have been on the fediverse for a few years, I've never done an #Introduction post, so let's do one for this channel.
I tend to post about actresses and singers I'm a fan of, sharing news about their new projects as well as my thoughts about films and tv I watch. In between I give my take on political things as well as about the fediverse. When I learn new things hosting a fediverse server or other technical things, I try to share them here too. Longer posts may end up on my blog gidikroon.eu (which you can follow at @GidiKroon@gidikroon.eu).
In real life I work as a software developer, but I keep work separate from a private account like this. I play (beach) volleyball when I can, at a very low level. I live
... show more
Hi, this me from my Friendica instance. I also post from @GidiKroon@mastodon.social, @gidi@pleroma.gidikroon.eu as well as from Twitter, Instagram and Tumblr as @GidiKroon but I'm mainly using my own Friendica for now. Even though I have been on the fediverse for a few years, I've never done an #Introduction post, so let's do one for this channel.
I tend to post about actresses and singers I'm a fan of, sharing news about their new projects as well as my thoughts about films and tv I watch. In between I give my take on political things as well as about the fediverse. When I learn new things hosting a fediverse server or other technical things, I try to share them here too. Longer posts may end up on my blog gidikroon.eu (which you can follow at @GidiKroon@gidikroon.eu).
In real life I work as a software developer, but I keep work separate from a private account like this. I play (beach) volleyball when I can, at a very low level. I live in the Netherlands and occasionally cycle and take photographs. I read and watch sci-fi and fantasy, and am interested in ways to achieve social justice, gender and race equality, sustainable energy use and climate justice.
Feel free to follow, as I will also follow other accounts I find interesting without expectation of them following back. Feel free to share (boost, repeat) my posts if you want, using whatever share function comes with your platform. I tend to close DMs to only allow people I follow. I check my fediverse accounts only one or a few times a day, typically on my phone or tablet, so I tend to not be part of live discussions.
You season 5 will be on Netflix on 24 April. I don't know which characters/actresses will come back, the two I'm hoping for I didn't spot in the teaser. #You #YouNetflix
New toy: a Laowa 2x macro lens. Comes very favourably out of reviews, also for non-macro distances. It's the crop sensor advantage that this is small-ish and cheap-ish, while still being a super-sharp f/2.8 apochromatic lens. Now I have to learn to manually focus, and to use (flash) lighting. #Laowa #Macro
Glad it's winter and there are no insects about, testing out my macro lens.
Canon R10 with Laowa 65mm F2.8, either with Godox speedlite flash off-camera or the light of my Samsung phone when I had already put the speedlite away.
There seem to be too many politicians at every level secretly thinking "Well the politics is rough and tumble, but at the end of the day we can shake hands and agree to disagree."
This kind of performative disagreement for the cameras is totally unsuited to governments that pose existential threats to their own citizens and the world in general.
I guess a lot of them were brought up in more gentle times and are trying to take their cues from that bygone age? Whatever the reason, they are going to let us all down if they carry on with their prioritising of civility above responsibility.
unicode proposal: you should be able to send an emoji with a specifier for which image set and version you want.
sometimes you want to send 😔, and you don't mind which version is displayed. but sometimes you want to send 😔, and have it specifically be the Samsung TouchWiz 7.0 (2016) version.
no other iteration of 😔 captures the feeling of "sigh... if only my girlfriend Chiaki Nanami from Danganronpa 2 was real... then i could be her Ultimate Gamer Husband instead of studying for this lame physics exam..." quite like this one
sometimes you want to send 😔, and you don't mind which version is displayed. but sometimes you want to send , and have it specifically be the Samsung TouchWiz 7.0 (2016) version.
@IngaLovinde yes but ideally every client would have to contain every iteration of every emoji set, so i could just send U+1F614 PENSIVE FACE followed by U+200D ZERO WIDTH JOINER and U+EA07 TOUCHWIZ SEVEN MODIFIER
> A #cosmic object spotted by the James Webb Space Telescope has flummoxed #astronomers. Now, a research team has studied hundreds of these "little red dots" and found clues about their identity.
And it's that resource usage for training that is my problem with it, so I'm not happy about AI subtitles in VLC. I also suspect it doesn't work, but that is maybe just me.
So, here is the first concrete technical info about VLC AI subtitles:
* it uses OpenAI whisper model. * you'll have to download that bit separately [1]. * vlc will use that on your computer, so it is not using the paid for API version.
So still using all the water and power to train models, and to run on your computer when you watch something that need translation.
VLC va intégrer de l'IA générative dans son lecteur vidéo open-source. Le plus célèbre des logiciels français prépare un générateur automatique de sous-titres capable de parler plusieurs dizaines de langues.
It’s about ChatGPT and whatnot innit. (Degraded text reading "You wouldn't steal the sum total of human knowledge and experience to use in your propri...
Why are people so proud of their religion and class and caste instead of, you know, being proud of themselves as a person? I'd think that's something a whole lot valuable.
The oldest surviving deck of 52 playing cards (1475 AD), made in the Netherlands 🇳🇱
The concept of playing cards is thought to have originated in Asia, making its way to Europe by the late 14th Century AD, likely through trading routes.
In the Netherlands, cards became popular during 15th Century AD, as both a pastime and a means of socialization among nobility and the growing merchant class.
Temple of Poseidon is one of the most important monuments from the Golden Age of Athens. It is perched on the edge of a cliff, located 70m above the sea, located at Cape Sounion 🇬🇷
Built in 5th Century BC, the temple is in the typical Greek hexastyle - featuring six doric columns on the front portico. 16 out of the original 38 columns are still standing today, and they were 20ft tall.
Whether you like the direction that things are going with AI or not you need to stop deluding yourself into thinking that you'll always be able to tell it's fake, you are going to eventually get scammed.
There are still some tells if you really know what you're looking for, but we're not at Will Smith spaghetti, 6 fingers or 2 left feet. Pretending like this is not a real problem you have to deal with isn't going to make the problem go away anymore.
UPDATE: I have created an account for updates and other information on FediMeteo - follow the account @admin to stay updated!
UPDATE: Ireland, Poland, Portugal and Switzerland have just been added
Weather has always influenced our lives: from agriculture to outdoor activities, to extreme events that, thanks to modern technology, can now be predicted with greater reliability. Personally, weather plays a significant role in my daily decisions, which is why I decided to create a service tailored for the Fediverse.
FediMeteo uses Open-Meteo data to publish updates every 6 hours, including current weather conditions, forecasts for the next 12 hours, and predictions for the upcoming days. Each country is served by its own dedicated instance (e.g., it.fedimeteo.com for Italy), managed through snac to ensure simplicity and efficiency in publishing.
You can follow FediMeteo directly in the Fediverse (on Mastodo
... show more
Announcing FediMeteo – Weather in the Fediverse!
UPDATE: I have created an account for updates and other information on FediMeteo - follow the account @admin to stay updated!
UPDATE: Ireland, Poland, Portugal and Switzerland have just been added
Weather has always influenced our lives: from agriculture to outdoor activities, to extreme events that, thanks to modern technology, can now be predicted with greater reliability. Personally, weather plays a significant role in my daily decisions, which is why I decided to create a service tailored for the Fediverse.
FediMeteo uses Open-Meteo data to publish updates every 6 hours, including current weather conditions, forecasts for the next 12 hours, and predictions for the upcoming days. Each country is served by its own dedicated instance (e.g., it.fedimeteo.com for Italy), managed through snac to ensure simplicity and efficiency in publishing.
You can follow FediMeteo directly in the Fediverse (on Mastodon and compatible platforms), via RSS, or by visiting the dedicated page for your city (e.g., fr.fedimeteo.com/paris).
Currently supported countries include: Austria, Germany, France, Ireland, Italy, Netherlands, Poland, Portugal, Spain, Switzerland and the United Kingdom, – with many more regions coming soon!
FediMeteo is hosted on a FreeBSD-based VPS, with each country isolated in its own jail to ensure security and scalability.
Visit the main site to explore the national instances and start following your local weather updates today: fedimeteo.com
Happy weather monitoring to all! 🌦️
FediMeteo is dedicated to my grandfather, who every evening would give me the weather forecast based on TV, radio, and his personal experience. He would convince me that the weather would be bad, so he had an excuse to accompany me to school instead of me going alone.
FediMeteo brings real-time weather updates to the Fediverse, covering cities across multiple countries. Explore country-specific weather forecasts and RSS feeds.
@FediThing oh, I thought almost all Irish people knew the Irish language. At least this has been the impression I had the last time I've been in Dublin. Do you think I'd better use English?
I guess you're going to run into this even more with Switzerland?
EDIT: Maybe there could be multiple accounts for countries with several languages? Would this be feasible, or would it add a lot of complication to FediMeteo?
@FediThing having multiple accounts for multiple languages could be problematic, also because this would raise the number of the API calls to the open meteo endpoint, which is limited. I'll think about it. I have the same problem for Switzerland, and the reason why it isn't there is that I couldn't decide which language to use there. After some comments from some Swiss users, I decided I'll use English.
@Bruticus Yes, I'd like to, but I need some more time. I'll probably use another VPS (located in the USA) and include the American continent. I'm still evaluating if it's better to create a single "usa." instance or divide or by state.
Might be worth doing by state as there is such a high level of repetition of place names in the US? Americans usually include the name of a state when describing a city for this reason. (The disambiguation page for Miami is a good example: en.wikipedia.org/wiki/Miami_(d… )
Trump promised "mass deportations". That will not mean deportations. Deporting people is not actually that easy. What it will mean is mass roundups and detentions. "Mass" means difficult to police the edges of. It means criteria will be arbitrary. Ethnicity-based by design, targeting people politically in practice.
There are no good and bad deportations. Someone's documentation status does not mean they deserve to be uprooted, detained in a desert, sent to some other country.
However, there is a *second* problem. In the previous Trump administration there was a running issue where Trump tried to deport people to Mexico, and Mexico did not allow this for a simple reason: It did not believe these people were Mexicans. And in fact, most of them probably were not.
To deport someone to a country, that country has to accept them. Some (most) of Trump's deportees will be nationals of some country other than Mexico, or citizens of the United States. Mexico has no particular reason to accept the deportation of a U.S. citizen to Mexico. The processing time caused by Mexico trying to figure out which of these non-Mexicans it will accept will mean even the Mexican nationals will be "detained" (imprisoned) for long periods while the backlog proceeds.
Again: No one deserves this. Being born in Mexico does not mean you should be sent to Mexico. But there's a bad problem and a worse problem and the worse problem is *what happens when the U.S. government holds large numbers of people its officers have already mentally dehumanized indefinitely in remote locations in Texas*.
The first thing that will happen is that the detainees will be subjected to forced labor. The second thing (or maybe first thing) is that some of them will die.
U.S. immigration is a never ending horror show and people not subject to its cruelty quickly stop caring. The "mass deport anyone who looks possibly Mexican via military tactics" thing has been done before, "Operation Wetback", 1954. Most people haven't heard of it. (And yeah, U.S. citizens got deported, if that matters.) People talk about Trump's walls and "camps". They talk less about Clinton, Bush, Obama and Biden's walls and detention camps. Biden opened new detention facilities.
When people stop paying attention, the horrors multiply. The greater the speed and density of events, the less our already-weakened journalistic system can notice any single one. We probably still don't know the worst thing that happened during the last Trump detention/deportation wave. The promise of the next one (that word, "mass") is simply that *the events will be denser*, and even the Trump cadre members don't care what the events will be. They just plan to do more.
When you study the third reich, you encounter this idea of "working toward the führer". Donald Trump is a lazy, incompetent and easily distracted man. But so was Adolf Hitler. (You can build a case one cause of the downfall of the nazis was Hitler inserting his approval as a requirement into military decisionmaking processes and then either making mediocre decisions or simply refusing to get up early enough in the morning, to give the required go-aheads early enough to be acted on.)
You can't predict the future. Trump administration II by its nature be unpredictable. It's dangerous to make a hard prediction (mine: "at least one U.S. citizen will be targeted for activism against the Trump administration, sent to an immigration detention camp, where they'll be detained indefinitely because there's nowhere to deport them, and eventually die there") because it could be wrong. Like so many of Trump's promises, the "mass deportation" idea might be simply forgotten.
But I don't think so. For two reasons. One, Trump has long had an interest in creating some sort of federalized police force that reports directly to him and can be deployed on arbitrary whims, and in the one test case of this ( npr.org/2020/07/17/892277592/f… ) ICE/CBP was in part operationalized to achieve it. To repeat: Customs and Border Patrol was deployed at least once to police a protest against Minneapolis police brutality, not a customs issue, and nothing legally stopped this then.
And the other reason is "working toward the führer". Trump administration I did not go as badly as it could have because Trump's underlings would simply ignore things he asked them to do, and he lacked the focus to follow up. I believe Trump admin II will be the reverse; this time he will have underlings in sync with his vile values from the start, and they will come up with and act on "Trump-like" things Trump didn't even think of, without Trump even being involved in the process.
Let's suppose Adolf Hitler was the most evil person who ever lived. The Hitler administration was more evil than even he himself was, because he had a fractious cabinet full of self-directed people, managing their own fiefdoms within the reich, *imagining* what Hitler would do and acting on that. And being, ultimately, much more creative and efficient at it than Adolf Hitler (again, a deeply mediocre man at best) himself could have ever been.
It's risky to project intentionality onto the actions of Donald Trump. But he seems to try to set up situations where people are encouraged to exercise their worst impulses. The CBP oddness in the Portland George Floyd protests seems to have come from a *hope* the CBP officers would cut loose and brutalize; instead they drove people around confusedly a bit then dropped them off. On Jan 6 2020 he encouraged people at a rally in Washington DC to do *something*, and they did… something.
The danger of Trump II is that Trump (a man too out of it to execute any particular plan himself) creates a space encouraging a cabinet member to be their worst selves who sends an underling admin to be their worst selves who sends a CBP officer to be their worst selves to one single person the CBP officer has mentally coded as an "immigrant" not a person, and this happens in a target-rich environment of many "immigrants" detained, arbitrarily and with no one caring, in remote Texas.
And there's a lot of things I'm afraid of. I'm afraid of Trump waking up one day and making a nonsense Presidential decree against trans people on Truth Social, which the Democrats and NYT liberals are too toothless and indifferent to notice or care it's illegal. I'm afraid the "war" in Gaza simply won't end, and will continue until no one is left alive. But what I can't stop imagining is the worst thing CBP/ICE could do with no one watching when you give them a pool of victims.
I don't want to write about this. I don't want to put these bytes on the Internet. I don't want to put these words in your head. I don't want this to be the reality we live in. I don't want to be right.
Sadly, you most likely are right. I believe there's another component to this, trump wants to be as famous as Hitler and for the same reasons, though never being defeated or seen as wrong.
On the gates of Auschwitz:
"Arbeit Macht Frei" Translation "work makes one free"
I kept trying to explain to cis het white people what I was scared of. I kept having a problem where whenever I made a specific prediction (even a prediction like "Trump will keep this widely publicized campaign promise"), they would come up with a reason why my specific prediction would not happen exactly as I laid it out, and therefore I have nothing to worry about.
The stuff you're saying is the stuff I'd been saying. It's refreshing to see someone else say it instead of dismiss it.
there's also the fact that many people will likely try to anticipate what trump wants based on his public statements, in hopes of currying favor in order to advance
the counter-argument to this is to say that in large parts of the country the people that would feel empowered by trump being president already feel empowered, and in the parts of the country where they're not already empowered (to the extent such places exist) they have to fight the state and local people that are also there on the ground
@bob I don't think this first part follows. If they feel empowered, they act. I think you're saying if they were going to do this they'd have done it already. But what we see is that *as* they do it, they realize they can do more, and so do. They did more under Biden than they did under Trump, simply because Biden happened after Trump. They are doing the things I am afraid of already. The camps are already here. The problem is acceleration and conditions that encourage acceleration.
@bob The difference between monstrous immigration policy under Trump and monstrous immigration policy under a hypothetical president Harris, as I see it, is the cap on how bad things can get.
right I think we're actually agreeing except I'm skeptical there would have actually been such a cap under Harris (Trump's monstrous immigration policy was an extension of Obama's monstrous immigration policy)
that (directly controlled police force and synergetic underlings) very much describes the ascend of Putin and his administration and all the horrors of it and i’m scared of this
@yottatsa There are a lot of scenarios which people are dismissing as entirely implausible, hysterical doomerism but which, factually speaking, either have been successfully deployed in the past or are currently being successfully deployed somewhere in the world right now. I don't know if it makes sense to be scared and I don't think it's good to panic. I don't find those emotions useful. But I don't see a counter-argument against the idea the worst case scenario could happen.
Apparently everybody totally forgot about the "kids in cages" horrors of his last time in office. I don't remember it even being mentioned during the election campaign. And that's utterly shameful. Everybody was just yakking about the economy and ignoring all the abhorrent shit he already did, and promised to do again. I guess America gets the president it deserves.
1/ On this last day of 2024, I thought this might interest those of you following our progress in detail. Here is a summary of all the One-Way Light Travel Times of the Voyager spacecraft over 2024.
The first thing to note is that the plot lines have different vertical axes! While the times are _offset_, the scale is the same. If you plot both these times on a single vertical scale, you can barely see the sinusoidal shape. Remember: green with green, blue with blue.
OK, why is there this shape? Because the distance measurements are being made from a moving platform: Earth. The sinusoidal variance is because of Earth's orbit around the Sun. At certain times of year, the Earth is actually moving towards the Voyagers faster than they are leaving, so the distance _decreases_.
Looks like Google did a not so nice thing. The Google Authenticator app no longer will generate the codes for some items, including Google's own account. This happened after they offered to backup all items to their cloud in case you lose your phone; if you don't do that, their app will break itself like this.
Two-factor authentication is about 'one thing you know and one thing you have'. The first thing is your password, the second thing your phone. The generated codes on your phone prove that you have your phone. Backing up the app in your Google account renders the latter useless and puts you back at one-factor authentication.
Of course you can and maybe should backup the secrets stored in the app. But that must be on 'something you have' and not somebody's cloud. #GoogleAuthenticator
Over six decades, Jill and I had the honor of calling Jimmy Carter a dear friend.
But, what’s extraordinary about Jimmy Carter, though, is that millions of people throughout America and the world who never met him thought of him as a dear friend as well.
As long as I live, I will never understand why America would rather elect a singularly worthless white rapist as president than an exceptionally impressive black woman.
Democrats were using tactics that are 50 yrs old and failed badly. I am embarrassed that we elected the sickest individual on the planet.. there is no coming back from this.
You have to tell most voters what they want to hear. There is a traditional American script. Americans want to hear that they are OK, not bad; hard work will be rewarded and with that & a little luck you can get to the top; even if you're down, you can pick up, dust off, & start again; & if there is anything wrong is the the fault of Others--foreigners, POCs, gays, druggies, layabouts, hippies; those who are 'not US.' The Black woman did not deliver that message. The rapist did.
Discover insightful articles on 4 Myths About Slavery We Should Stop Believing Now. Join us in exploring solutions for a just, sustainable, and compassionate world. #4 Myths About Slavery We Should Stop Believing Now
She prioritised corporate interests over citizens.
She backs the genocide. She gaslit those who speak out of protest against it. She banned Palestinian voices at her DNC, where they openly mocked deaf children.
money, as usual. They want it and he convinced them the reason they don’t have it is because of people like her. It allowed them to not feel bad about failing, they weren’t not good enough the game just wasn’t fair. He took away their guilt and gave them hope for the future, that’s why he won. It was a lie but it still worked.
Rape is a patriarchal show of force, and thus status-giving. Worthless doesn’t matter because patriarchal status is much more valuable. White is higher in the American patriarchal hierarchy. Man is obviously higher in a patriarchal hierarchy.
He also displays a lot of things that are strengthening patriarchy, like anti-trans and anti-science.
Patriarchy might not explain everything about why he won, though.
Unfortunately, it is simple: About 5 million US voters DO NOT think that a woman can be president (See also Hillary's vote). I counted about a 3 million person drop in Democratic vote each time the Dems proposed a woman. Sigh.
The poles leading up to the vote said it: Who will do a better job for Economy or Foreign affairs? No one asked "WHY THE FXXX do you think that a failure at running SUCCESSFUL businesses would do better?
because they cared so much about Palestinian genocide - see they are flying in droves in region to protect them right - led by Jill Stein - yeah right, and because the thsi booming economy is not helping the ones at the bottom of the pyramid - so the only way to fix it is by sending even more to the top. Nothing at all to do with protecting their existing privileges, wealth and worse their belief in the same thing as the felons side but just cowardly enough to go vote for him.
One can speculate, people voted for the rapist felon with 77 million votes over the DNC, by 2 million votes with 75 million votes. So 90 million did not vote and why? Apathy? Tired of voting for the best of the worst? DNC has a history of playing leap frog with the GOP so nothing changes? Australia has mandatory voting so a majority wins, which is more democratic, instead of 45 wining with less than a 1/3 of the eligible voters? Any ideas why 90 people million did not vote?
The deep historical forces that explain Trump’s win
The research team I lead studies cycles of political integration and disintegration over the past 5,000 years. We have found that societies, organised as states, can experience significant periods of peace and stability lasting, roughly, a century or so.
they then enter periods of social unrest and political breakdown. e.g. the Roman empire, the English civil war ... they slid into crisis, and then emerged from it.
Our research shows that political breakdown, from the Roman Empire to the Russian revolution, follows a clear pattern: workers’ wages stagnate, while elites multiply
three "lesser of two evils" elections eroded our standards to the point that we were expected to just avoid mentioning the ongoing genocide biden and kamala were sponsoring. in the end it was just "fuck you, i'm not the other guy, that's enough", and being vaguely respectable.
even with that, it was one of the slimmest margins in a long time, and trump didn't even win half the popular vote.
kamala could have won. america didn't choose trump, they stopped choosing democrats.
i think the people that abstained made a very bad choice. i doubt the democratic party will learn from their mistakes, or even care as much as we do that they lost. but the democrats and the abstaining are both to blame.
yes, we agree with your description of events, and with your prediction
we do also think that a critical factor here is that a lot of people have come to see politics as more of a spectator sport, and less as something with real consequences
My understanding is simply misinformation and DEEP divide between the information both sides got to see. My parents have a fundamentally different understanding of the political scene than I do. We aren't white and are immigrants but they voted for him because they don't believe he would actually deport citizens. What I know of Trump is so utterly ridiculous that it's just more believable that democrats did some underhanded politicking and media control.
Any answer to the question that does not place the responsibility fully on the people with actual political power is wrong and will lead to Democrats losing yet again. Blaming the powerless is easy and popular and makes people feel superior, but it's just scapegoating.
Why? Well, Abraham Lincoln supposedly said, "You can fool all of the people some of the time and some of the people all of the time, but you can not fool all of the people all of the time." An obvious corollary is that sometimes you can fool enough of the people for enough time to get elected and then it just might be too late.
My dishwasher was like 'hold my beer' and didn't hold its water anymore. Meanwhile my tablet seems to be fixed for now by uninstalling the TikTok app (it literally caused the display to flash, have interference, etc).
Insider trading tip for people trading in Samsung shares: both my tablet and phone are showing near-end-of-life signs and I may soon need to be buying new ones. With the insane device prices nowadays, this sure is going to raise their shares.
(Just to be clear, the above is obviously sarcasm. I have no inside knowledge. Just a consumer.)
Insider trading tip for people trading in Samsung shares: both my tablet and phone are showing near-end-of-life signs and I may soon need to be buying new ones. With the insane device prices nowadays, this sure is going to raise their shares.
(Just to be clear, the above is obviously sarcasm. I have no inside knowledge. Just a consumer.)
I finally turned off GitHub Copilot yesterday. I’ve been using it for about a year on the ‘free for open-source maintainers’ tier. I was skeptical but didn’t want to dismiss it without a fair trial.
It has cost me more time than it has saved. It lets me type faster, which has been useful when writing tests where I’m testing a variety of permutations of an API to check error handling for all of the conditions.
I can recall three places where it has introduced bugs that took me more time to to debug than the total time saving:
The first was something that initially impressed me. I pasted the prose description of how to communicate with an Ethernet MAC into a comment and then wrote some method prototypes. It autocompleted the bodies. All very plausible looking. Only it managed to flip a bit in the MDIO read and write register commands. MDIO is basically a multiplexing system. You have two device registers exposed, one sets the command (read or write a specific internal register) and the other is the value. It got the read and write the
... show more
I finally turned off GitHub Copilot yesterday. I’ve been using it for about a year on the ‘free for open-source maintainers’ tier. I was skeptical but didn’t want to dismiss it without a fair trial.
It has cost me more time than it has saved. It lets me type faster, which has been useful when writing tests where I’m testing a variety of permutations of an API to check error handling for all of the conditions.
I can recall three places where it has introduced bugs that took me more time to to debug than the total time saving:
The first was something that initially impressed me. I pasted the prose description of how to communicate with an Ethernet MAC into a comment and then wrote some method prototypes. It autocompleted the bodies. All very plausible looking. Only it managed to flip a bit in the MDIO read and write register commands. MDIO is basically a multiplexing system. You have two device registers exposed, one sets the command (read or write a specific internal register) and the other is the value. It got the read and write the wrong way around, so when I thought I was writing a value, I was actually reading. When I thought I was reading, I was actually seeing the value in the last register I thought I had written. It took two of us over a day to debug this. The fix was simple, but the bug was in the middle of correct-looking code. If I’d manually transcribed the command from the data sheet, I would not have got this wrong because I’d have triple checked it.
Another case it had inverted the condition in an if statement inside an error-handling path. The error handling was a rare case and was asymmetric. Hitting the if case when you wanted the else case was okay but the converse was not. Lots of debugging. I learned from this to read the generated code more carefully, but that increased cognitive load and eliminated most of the benefit. Typing code is not the bottleneck and if I have to think about what I want and then read carefully to check it really is what I want, I am slower.
Most recently, I was writing a simple binary search and insertion-deletion operations for a sorted array. I assumed that this was something that had hundreds of examples in the training data and so would be fine. It had all sorts of corner-case bugs. I eventually gave up fixing them and rewrote the code from scratch.
Last week I did some work on a remote machine where I hadn’t set up Copilot and I felt much more productive. Autocomplete was either correct or not present, so I was spending more time thinking about what to write. I don’t entirely trust this kind of subjective judgement, but it was a data point. Around the same time I wrote some code without clangd set up and that really hurt. It turns out I really rely on AST-aware completion to explore APIs. I had to look up more things in the documentation. Copilot was never good for this because it would just bullshit APIs, so something showing up in autocomplete didn’t mean it was real. This would be improved by using a feedback system to require autocomplete outputs to type check, but then they would take much longer to create (probably at least a 10x increase in LLM compute time) and wouldn’t complete fragments, so I don’t see a good path to being able to do this without tight coupling to the LSP server and possibly not even then.
Yesterday I was writing bits of the CHERIoT Programmers’ Guide and it kept autocompleting text in a different writing style, some of which was obviously plagiarised (when I’m describing precisely how to implement a specific, and not very common, lock type with a futex and the autocomplete is a paragraph of text with a lot of detail, I’m confident you don’t have more than one or two examples of that in the training set). It was distracting and annoying. I wrote much faster after turning it off.
So, after giving it a fair try, I have concluded that it is both a net decrease in productivity and probably an increase in legal liability.
Discussions I am not interested in having:
You are holding it wrong. Using Copilot with this magic config setting / prompt tweak makes it better. At its absolute best, it was a small productivity increase, if it needs more effort to use, that will be offset.
This other LLM is much better. I don’t care. The costs of the bullshitting far outweighed the benefits when it worked, to be better it would have to not bullshit, and that’s not something LLMs can do.
It’s great for boilerplate! No. APIs that require every user to write the same code are broken. Fix them, don’t fill the world with more code using them that will need fixing when the APIs change.
Don’t use LLMs for autocomplete, use them for dialogues about the code. Tried that. It’s worse than a rubber duck, which at least knows to stay silent when it doesn’t know what it’s talking about.
The one place Copilot was vaguely useful was hinting at missing abstractions (if it can autocomplete big chunks then my APIs required too much boilerplate and needed better abstractions). The place I thought it might be useful was spotting inconsistent API names and parameter orders but it was actually very bad at this (presumably because of the way it tokenises identifiers?). With a load of examples with consistent names, it would suggest things that didn't match the convention. After using three APIs that all passed the same parameters in the same order, it would suggest flipping the order for the fourth.
thanks for this write-up. there needs to be a broader societal understanding of how these things mess up, and of what happens when they do.
everything you wrote is something we'd have predicted, but we had specific knowledge about how the model works and the cognitive biases people have about it, which helped us get there. us saying it in the abstract doesn't really help anyone, people need real experiences to learn from.
@erincandescent my opinion on this is: you probably can, but in some cases that could require raising the abstraction level too much. DRY taken to the extreme is premature abstraction
your experience is how I expected mine to be if I had actually given the technology a chance.
Machine learning has been useful for decades, mostly quietly. The red flag against LLMs for me (aside from the authorship laundering and mass poaching of content) was there scramble by all companies to shoehorn it into their products, like a solution looking for a problem; I’ve yet to see it actually solve.
@carbontwelve I used machine learning in my PhD. The use case there was data prefetching. This was an ideal task for ML, because the benefits of a correct answer were high and the cost of an incorrect answer were low. In the worst case, your prefetching evicts something from cache that you need later, but a 60% accuracy in predictions is a big overall improvement.
Programming is the opposite. The benefits of being able to generate correct code faster 80% of the time are small but the costs of generating incorrect code even 1% of the time are high. The entire shift-left movement is about finding and preventing bugs earlier.
I try to scroll through photos shared on the Fediverse a few times a day (obviously, not all of them). It’s amazing to see the incredible work photographers are creating - not just the pros, but everyone. So many of these shots are true masterpieces. Browsing through them is a relaxing experience and makes me incredibly proud of all of you. It’s awesome that you share your work instead of letting it sit hidden away. Your photos have the power to inspire someone out there! vernissage.photos/editors
Beetlejuice Beetlejuice with Jenna Ortega and The Pod Generation with Emilia Clarke arrived on blu-ray!
The German release of the latter has an alternative title Baby To Go, which isn't even in German. In my own country hardly any discs are released anymore, so we have to make do with these German releases. #JennaOrtega #BeetlejuiceBeetlejuice #EmiliaClarke #ThePodGeneration #BabyToGo
Gidi Kroon
in reply to Gidi Kroon • •Don't handhold macro shots they say. Don't use the built-in flash and don't use maximum aperture. They are probably right.
(This is nowhere near the maximum magnification btw)
eabolden likes this.
Gidi Kroon
in reply to Gidi Kroon • •Glad it's winter and there are no insects about, testing out my macro lens.
Canon R10 with Laowa 65mm F2.8, either with Godox speedlite flash off-camera or the light of my Samsung phone when I had already put the speedlite away.
#Photography #macro #Laowa #Canon #Godox #Samsung
Andrew "Ace" Arsenault likes this.
Gidi Kroon
in reply to Gidi Kroon • •#Photography #MacroPhotography #Canon #Laowa #Godox
like this
Radgryd, tomo, kumazemi and Macuahuitl like this.