taylor

@btp@fosstodon.org Nice! That's an awesome haul!

@freediverx@mastodon.social @boojum@hachyderm.io @skinnylatte@hachyderm.io That they will protect your privacy to the limit that they legally can, perhaps? A privacy-focused email provider that doesn't abide by the laws of the country they operate in stops existing pretty quickly.

@Justin@social.younger.wtf Mmmm, spaghetti.

@RandamuMaki @zekjur The right solution would be for git format-patch and git am to use MIME multipart/mixed and deterministically and unambiguously separate the patch from the commit message. They're already generating an mbox format as it is.

@Setcab @zekjur Well, for patch it's a simple and useful tool. You get a diff and can apply the patch to the same code to the get the same result, and the user can write little comments between diffs to annotate them. The problem is that the format is not robust to arbitrary text between patch hunks, and git didn't use proper delimitation between the commit message and the patch. Ideally, the tooling would properly delimit the two, and also would never run the patch tool over the commit message as well as the actual patch. I think the format is designed fine for what it's meant to do, it's just being used poorly.

@imp3tuz Hey, thanks for the heads up! I just started a project that this will be extremely useful for. I'll definitely bookmark this for the future.

@InternetEh Honestly, I could understand their point of view if they were at least consistent. "Hard on crime" except for the pardons of actually guilty (but rich) criminal fraudsters. "The law matters" except when my politicians are violating the law. "The constitution is important" with the same exceptions. They care about legality only as far as they've been told to, otherwise it's considered a bureaucratic impediment. I think it's why right wingers are so successful lately. The left tends to have no qualms eating their own in purity spirals for almost nothing, while the modern right is basically immune to cognitive dissonance and has no trouble uniting behind somebody (even if they are the opposite of what they say their ethics demand), changing their own values to fit any new ones on a whim. Whatever problems the right has, they are easy to unite, trivial to rally, and impossible to sway otherwise in terms of logic and reason.

@beecycling @TheBreadmonkey Yeah, that's what I've been seeing with some of the LLM programming anecdotes lately (and some personal experience with work projects, where we're trialing this stuff). AI frequently makes fewer mistakes than humans, but the mistakes that they do make are very often bonkers and baffling from a human perspective, and can be a lot more dangerous as a result, because you can miss nefarious issues, being trained to account for regular human mistakes and stupidity.

@Splig Not bad at all! You could throw some flat materials on it and it would look fine enough as low-poly art in a few minutes without any real texturing work.

I liked this piece by @sako@mastodon.gamedev.place , so I recreated it in Blender with Geometry nodes. I spent a while mostly doing the Instance Grid node group. I got a simple version working pretty easily at first, but wanted something more robust, with functionality for edge margins, different margin types, margin collapsing, etc. I failed to just hack at that for a while before whiteboarding it out, which I probably should have done in the first place. Here's the blend file. #Blender #GeometryNodes #MoGraph #MotionGraphics #Looping #Loop #AbstractArt #3DArt

@btp@fosstodon.org That's awesome, man! I'm happy to hear it. You got any photos to share?

@btp@fosstodon.org What about just pictures of your signed stuff?

@paco@infosec.exchange @trashwizard@mastodon.social @Em0nM4stodon@infosec.exchange The energy use argument of AI isn't the gotcha you think it is. The average single prompt to Google Gemini takes about 0.24 Wh of electricity, which is the same as running a microwave oven for 1 second. If you are comparing it to how much energy it takes to run internet searches and scour forums for however much time, often the AI will come up to using less electricity overall. AI's environmental impact is huge, but each individual use is not nearly as huge, and shaming somebody for quickly finding an answer that they tried and couldn't find otherwise isn't productive, logical, or accurate in this case. Of course, using these increases demand for them, which necessitates training new models, which takes significantly more energy than prompting them, so it's not cut-and-dry, but that's my point, it's not cut-and-dry. This was personally useful for you and basically no one else. This is true every time somebody solves a problem they had and doesn't publish the results, and the energy use of solving the problem is still duplicated among everybody who has to solve the problem with or without AI. I'm not an AI booster or anything, but this line of dishonest argument is tired on both sides (the proponents are significantly more annoying about it than the detractors, to be fair). If they solved the energy use overnight, or built fields of solar panels to completely offset the energy use, would you be 100% OK with this AI use? Would most of the people complaining about the energy footprint be OK with it? With many of the artists complaining about copyright violation, would they be OK if one was trained on public domain work and was still stealing their jobs? AI is a symptom and an accelerator of what is broken about our economic systems, and the capitalist oligarchy. It's not the problem in and of itself, it's a magnifying glass on the fact that the game is rigged, and we don't have the structures set up to actually keep people fed and housed who need to be fed and housed.

@paco@infosec.exchange @trashwizard@mastodon.social @Em0nM4stodon@infosec.exchange https://arxiv.org/abs/2508.15734 Our approach accounts for the full stack of AI serving infrastructure - including active AI accelerator power, host system energy, idle machine capacity, and data center energy overhead. Through detailed instrumentation of Google's AI infrastructure for serving the Gemini AI assistant, we find the median Gemini Apps text prompt consumes 0.24 Wh of energy - a figure substantially lower than many public estimates. We also show that Google's software efficiency efforts and clean energy procurement have driven a 33x reduction in energy consumption and a 44x reduction in carbon footprint for the median Gemini Apps text prompt over one year. We identify that the median Gemini Apps text prompt uses less energy than watching nine seconds of television (0.24 Wh) and consumes the equivalent of five drops of water (0.26 mL). While these impacts are low compared to other daily activities, reducing the environmental impact of AI serving continues to warrant important attention. I never said that power consumption is a nothingburger, or anything of the sort. I care strongly about environmental impact and think that the energy use of AI is problematic and definitely needs addressing. The thing is that they are actively being addressed, and we're rapidly approaching a situation where the environmental ethics of asking a LLM something is closer to the moral question of whether you should heat up your canned food or eat it cold. If you can measure heating up a can of beans in terms of hundreds of LLM prompts, why harass individual people about the energy use of these prompts? The energy use is important, but I'm much more concerned about the social, political, and economic effects of AI than the energy use at this point. The average person has many more impactful areas that they could cut their energy use (or direct fossil fuel use) that rank far above asking a LLM a question now and then. Personally, I'm selfishly hoping (but not expecting) that the LLMs drive the construction of tons of solar farms, then the VC money dries up, the AI bubble pops, demand drops hardcore, and suddenly we have an overabundance of green energy that demolishes fossil fuel need entirely. Edit: Also, I regret and apologize for calling your reasoning "dishonest". That was kneejerk, and it wasn't warranted. I'm personally quite critical of the AI industry and much of AI use, and I tend to react harshly when somebody whose position I generally agree with is using reasoning I consider weak or fallacious. It feels like it undermines my own position, and gives ammunition to the opposition. That's a me problem.

@Ronflaix@mastodon.gamedev.place @NullTheFool@mastodon.gamedev.place you can still pass anything other than the type you provided, it'll just cause a few warnings but no performance or codegen change Note: Python will not give any warnings, but other tooling (like IDEs and LSPs) will.

@fedilore@mastodon.social How would that work for recurring payments? For one-off payments, sure, you can take payment, mark the account as paid for a time period, and not keep that connection, but for a subscription, the card has to be associated with the account to be scheduled to pay it in the future.

@freediverx@mastodon.social @boojum@hachyderm.io @skinnylatte@hachyderm.io How do you charge somebody's credit card regularly on a subscription basis without storing any PII or working with a third party that stores PII? This is like charging a premium for the world’s most secure door lock when a burglar can easily enter your house by breaking a window. In which case it would be ridiculous to blame the door lock company when somebody breaks your window. I think a more apt analogy is that it's like a hotel charging a premium for a high-privacy room, but still giving their card records to the FBI when they come with a warrant for those records, telling them what credit card paid for which room.