Zimmie

@analog_feelings @WiteWulf @nina_kali_nina In that FreeBSD pulls from upstreams, I guess. They specifically call out “AI” emitted code as the kind of low-quality stuff they don’t accept: https://mastodon.social/@allanjude/116320464321208949

@cabbagebeets @notjustbikes New nuclear plants probably shouldn’t be built in most areas (people living close to or inside the arctic/antarctic circles may not have better options). Renewables are worthwhile, and should absolutely make up most new power generation. That said, all power generation has waste, and renewables have waste problems of their own which should not be ignored. We don’t have cost-effective recycling for solar panels. When panels break today (hail, high winds, sand etching the surface over time, etc.), it’s much cheaper to dump them in e-waste landfills in Africa or Southeast Asia, where they poison the groundwater. Better recycling is on the horizon (How long have better nuclear reactors been on the horizon?), but it’s not here yet. Legal changes like making recycling mandatory could help, but most areas don’t seem to have the stomach for that. Wind turbine blades are composite materials which probably can’t ever be recycled. The best disposal method we have for them today is grinding them up (which risks giving workers silicosis; see quartz countertop issues) and burning them for industrial process heat.

@cabbagebeets @notjustbikes I’d be happy to *eat* all of the nuclear fuel spent to generate the power I personally use in a year. People really have no concept of how energy-dense it is. This is an aerial photo of the storage for Maine Yankee Nuclear Power Plant, a plant which operated from 1972-1997, rated at 860 MW, with a lifetime capacity factor of 68.2% (so it generated ~586 MW on average). 25 years is 219,150 hours, so this plant generated ~128.5 TWh of power over its life. And this isn’t just spent fuel, it’s all the severely contaminated concrete and steel from the reactor, too. All in a grid about 100’ by 120’. Last I heard, New York City’s power consumption is estimated at 60 TWh per year, so this picture represents all the waste from powering one of the biggest cities in the world for two entire years.

@Viss @mroach @FritzAdalis @jschauma Or “Terrapin”. An attacker has to be in a position to discard arbitrary packets from an SSH negotiation, and the only impact on most versions of OpenSSH is a DoS. Which an attacker in that position could cause by discarding the SYNs. Better spend person-months in change control to disable chacha20-poly1305 everywhere!

@violetmadder @TeflonTrout @hellomiakoda @craignicol @aesthr @mercuryjohn That’s a common explanation, but it’s usually a bit incorrect. Stations have extremely variable sizes. Some of the bigger ones in growing suburbs could handle a two-alarm response solo (and may have the necessary support staff in the station). Lots of stations (especially in dense cities, where ground-level space is at a premium) have one or two engines and a single ladder truck, so four or five such stations might respond to a two-alarm. Ultimately, though, the point is “alarms” is a linear-ish scale of response size which isn’t well-standardized between areas.

@hellomiakoda @craignicol @aesthr @mercuryjohn The exact meaning varies between regions, but an “alarm” is a measure of response size. Each is usually ~4 engines, ~2 ladder trucks, and 3-5 people each. Two-alarm is ~8 engines and ~5 ladder trucks, plus a bunch of support staff (rehab unit to help firefighters cool down and hydrate, safety chief, tactical support, communications unit, etc.). Three alarm adds another 4 engines, 2 ladders, and some more specialized support staff. Four alarm adds another 4 and 2, plus planning specialists. Five alarm adds another 4 and 2. This is 30-35 total trucks and ~130 firefighters plus support staff.

@vees I’m partial to “Financial crimes reporting, this is an unsecured line.”

@only_ohm @fesshole Depends on where in training. Phlebotomists (and EMTs, paramedics, etc.) learn to place needles in veins by sticking each other in class. I’m sure the majority of my classmates would have been up for our blood being used for something like this.

@simonzerafa @dpk Depends on the variant of A14/M1. The M1 in the MacBook Air, early MacBook Pros, Mac mini, and iMac is 4p4e7g or 4p4e8g. The A18/M4 in the MacBook Neo is 2p4e5g, but with much newer performance cores. It beats the base M1 in single-core, multi-core, and Metal performance. The M1 Pro beats it in everything but single-core performance.

@aud Of course, when the actual goal is to victimize kids, both the laws and the pervert glasses make sense.

@thomastraynor @dangoodin At least some ad-blocker-blockers are shown by default, then disabled by something loaded with the ads. If you don’t load the bypass from Doubleclick or wherever, the script runs and shows the “Home taping is killing the record industry” popover.

@mwl@io.mwl.io

@MissConstrue @dwm @cstross Take a look at the tiny power cord. Sure, it’s uselessly short, but I more mean how thin it is. There’s no way I would trust that to carry even three Amps, so the whole device could do maybe 300W. Split between 79 total sockets (66 120V, 13 USB ports), that’s not even 4W per socket before the power strip is likely to melt its upstream wiring.

@cstross@wandering.shop @cm@chaos.social The newer simulated top-down view provided by side cameras and a front camera is also nice for checking whether you need to realign inside a parking space before you stop the engine. Hit the bottom, confirm I’m centered, shut down.

@pier @shredder7579 @martin @jk Yeah, I just gave the layout constraints as an example of the boilerplate people would love to optimize away. It’s verbose, and similar enough to make people go cross-eyed checking it, but just different enough each time it’s hard to make a macro or whatever. Exactly the kind of stuff LLMs are pitched as handling well. And they simply don’t. All the unethical behavior, resource waste, and so on, and the result is deeply mediocre. It’s weird that *anybody* supports it.

@pier @shredder7579 @martin @jk I decided to try out some code generation to see how it works lately. To do this, I built a small, composite view in AppKit (the macOS UI framework). In my case, this involves subclassing NSView, then setting up my subviews in the initializer. AppKit uses “constraints” to specify how views should be arranged relative to their container and to other views within the container. When laying out views in code, these are generally a *long* line to specify each of top, bottom, leading, and trailing. I built my first subview, and the model recommended adding the constraints for that view. The recommendations were all correct. Incidentally, they looked exactly like constraint sets I had written for views elsewhere in the application (same order, style, etc.). I was impressed. Then I added a second subview, and it recommended some incorrect constraints. I went on for a while, but it never got better than ~30% correct recommendations. It recommended variable names which don’t exist, methods which don’t exist, incorrect property names and parameter names, logic errors, all kinds of stuff. Definitely not worth the resources being thrown at it.