you’re just not going to get working men to cook healthy meals for themselves.
When I was young the answer to this was "steak". Quick and easy to make a good one, gentle learning curve to make a great one, feels manly to cook, pairs with any sides, great for protein+iron.
But although I'm trying to hold back from adding to the "$20 burrito" discourse (with at least two links and a page of ranting), I have to admit that steaks actually are becoming steadily more expensive, even adjusted for wages.
It almost seems like the only men doing this are into bodybuilding
Right? If you can talk yourself into going all-in on the protein power of lean chicken, that's still affordable and also a decent meat with which to practice cooking skills, but making chicken (at least with a stove top or oven; grilling every day can be a hassle) instead of beef you just don't get the same "primally in touch with my caveman ancestors", "top of the food chain over mammals 5 times my size", "nature red in tooth and claw and steak knife" vibes.
Shrimp have become cheaper over my adult lifetime even in nominal dollars, gone from luxury to staple prices in real terms, but that might be even trickier these days? IMO making fancy scampi for a date feels manly, but for young men self-conscious about their masculinity I'd worry that the "I will not eat the bugs" meme may deter them from literal arthropods.
roystgnr
13hr ago
I'd agree with all of this, but the one thing that you omitted that raises my hackles with Pangram is that (at least in all the screenshots I've seen) they output a point estimate rather than a range. You've got a test that inherently is going to have at least a few false positives and a false negatives (they claim 1-in-10k, and report third-party tests that suggest around 1-in-40 false negatives and N-was-too-low-to-find-one false positives), but you can't give us a 95% CI instead of just the mode (mean would never hit 100%) of your estimate, to make it easier for us to be wary when we input a hard-to-classify passage? That does not signal respect for avoiding error.
roystgnr
13hr ago
I mostly use LLMs (including Gemini!) for math and software. I think by far the biggest third category of use cases for me is "better search engine":
- Finding any existing name(s) for a concept I describe.
- Finding old quotes, jokes, short stories, TV episodes, or historical events, when I only remember a description but not searchable exact words.
- Finding TTRPG rules and DMing suggestions for tricky situations.
- Locating data for particular statistics
- Getting quick but detailed summaries when I'm curious about music/history/physics/sociology questions
- Checking food safety rules, stain removal tips, musical instrument storage requirements, and injury care and pet care guidelines
- Finding online course options for my kids
- Planning trips
- Locating that one canyon my parents took me to 40 years ago, based on a vague description, so I could take my kids
- Finding fun "little" places to stop on the between-major-attraction legs of road trips that might otherwise be slogs
- Finding hiking trails that match my kids' preferences, or in general searching destination options for amenities that you can't find hard-coded filters for
- Learning "gotchas" about travel destinations ahead of time
- Coming up with packing/prep checklists
- Finding IRS rules based on a vague description of a situation
- Quickly finding information on a list of professors based on a page with just photos
Including "much better reverse-image-search":
- Identifying a foster kitten's skin problem by photo, and another cat's stomach troubles by description.
- Identifying some "volunteer" garden plants by photo, and some animals by photo of their droppings.
- Identifying garden pests by photo and figuring out how to safely get rid of them
- Identifying strange things we saw on vacation, from photos
And some miscellaneous:
- Figuring out how to better use my garden peppers
- As a better thesaurus, refining synonym/antonym lists by ruling out or requiring particular connotations
- Generative tasks
- Making funny pictures and videos with my kids
- Coming up with more interesting quiz problem ideas and game ideas for the math team I was coaching
- Distilling long news pages into short summaries of the most important bits
- For "handyman" problems
- Figuring out how to work around problems with a hydroponics pump wearing out, then how to replace it with a more efficient setup when it finally broke
- Figuring out to fix a broken tub/shower diverter valve
- Figuring out how to fix a broken smartwatch
- Troubleshooting a failing car ignition
- Translating photos of foreign-language signs/instructions/machines, and figuring out how to use some of those machines
YMMV, of course, but there are skill issues for everyone here. One ironic skill issue of my own that only occurs to me belatedly: I probably spent 5 minutes skimming logs and typing up this list by hand, but if I instead give Gemini a 3-sentence prompt and do something else for 15 seconds, I get a list that's arguably a little better despite being more abbreviated.
AFAIK there weren't even minor beatings for my father or his five siblings, but they're not necessary; reportedly "you don't get seconds unless you finish before there's nothing left, and no there are no snacks in the house" worked well enough to get everybody to generally eat what they were served, and quickly.
I don't do this myself and can't recommend this as a parenting style unless you're poor enough that food wastage is serious business, though. It doesn't actually get the kid to internalize healthy eating, and it might be a bit psychologically scarring. For the rest of his life my dad was always first in any line for food, and yet his idea of "vegetables" worth selecting voluntarily never expanded past corn and iceberg lettuce. His grandkids vary but even the pickiest isn't nearly that bad.
roystgnr
1d ago
and bikinis weren’t thé norm
¿Por qué tienes ese acento gráfico?
Anyway, anyone who remembers bikinis being normalized isn't middle-aged anymore; that mostly happened in the 1960s. The famous old-fashioned-style song, about even shy girls getting used to bikinis, hit #1 on the charts! Of the subsequent wave of 60s' "beach party movies" only a large minority had bikini in the title but most of the rest at least had bikini-clad characters, usually on the movie posters too. Bikinis weren't the norm in the sense that more than 50% of women wore them, but they still aren't: most women wear one-piece swimsuits still, and most of the rest wear tankinis.
Social media and Hollywood might be accelerating cultural change today, perhaps even relative to the 60s, but nobody's making their multi-billion-dollar datacenter investments because they need extra servers for Instagram clips or renderfarms for Disney movies, they're doing it because they really think better-trained bots can take a lot of jobs and/or immanentize the eschaton. If rural people think otherwise (citation needed?), I guess that's less of a stretch than the water use panic, but it's still kind of missing the point.
roystgnr
1d ago
an open letter [...] the letter [...] it
You'd think these phrases would be grammatically correct, but one mildly interesting bit of this story that your summary underemphasized is that technically they're not; "the letter" here doesn't describe a singular thing but rather a plural. The Good Law Project made a couple stealth edits to "the" letter after publishing and getting hundreds of online signatures with the first version of it. "Both conducted thorough investigations and found no evidence whatsoever of wrongdoing." (already with 600+ signatures in that archive snapshot) became "Investigations have found no evidence whatsoever of wrongdoing.", and then in addition to removing Professor Uchegbu's name entirely from the principal signatories they've deleted Lord Simon Wooley's affiliation with Cambridge for some reason, despite leaving it for three other Cambridge names there.
roystgnr
1d ago
This is good news for the Republican Party, and the primary reason I supported him.
Ah, yes, the "Pied Piper strategy". Still popular on the other side of the aisle too, arguably with some success, but I'd personally still try to steer clear of a strategy that's got an obvious and already-demonstrated failure mode for its users: sometimes even when you're sure the median voter will hate a guy as much as you do he simply wins anyway.
roystgnr
1d ago
My expectations for activities engaged in by arbitrary people are looser than my expectations for my friends', which are looser than my expectations for my wife's and kids', which are looser than my expectations for things I do together with my kids or my friends, which are looser than my expectations for things I do with my wife. None of these distinctions is a contradiction; they're the proper liberal way to thread the needle between "nobody should care about my preferences" self-abnegation versus "everybody should care about my preferences all the time" totalitarian attitudes.
It's totally reasonable to say that it's fine for a show to be made, either fine or not fine for your SO to want to watch it themselves, and yet not fine for your SO to pick it for the both of you to watch together. My wife likes phở bò with tripe and I like sushi with wasabi, and that's all great, but neither of us would ever order our favorite for the other one too, and if we did make a similar mistake out of ignorance then we'd talk about it, not to say that one of us thinks the other is gross, but to clarify what about the food we think is gross to us (e.g. I still like phở without tripe, she still likes seared tuna without wasabi) so a similar mistake doesn't happen again.
roystgnr
2d ago
Spoiler warnings! I've tried to read Project Lawful a few times now, and although I keep failing (IMHO it would be awesome with a skilled and bloodthirsty editor, but it'll never get one) I make it another 20+% through each time, and I'm still hoping to finish it someday without cracking and asking an LLM to take that editing job first.
roystgnr
2d ago
In the best set of current benchmarks I could quickly find, there are only four with leaderboard results for both Grok and Gemini, but Gemini is leading in 3 out of 4... which is definitely an underestimate; there are no leaderboards with Grok results but no Gemini results, and spot checking shows there's at least some Gemini leaderboard results where Grok was tested but no version made the leaderboard.
Gemini is not as competitive versus GPT (7 out of 17) or Claude (5 out of 16) though.
Subjectively it's pretty nice. I use Gemini to "self-" study advanced math (I'm a PhD, but applied math that leaned heavily into analysis, so my topology is weaker and my abstract algebra is pathetic), and even with its moderate/default setting I catch it in errors only about as often as it catches me, and its errors are probably technicalities more often than mine are. The other day I had it actually make an embarrassing error and double down, the first time it's done either since I started using it, but even then I just had to turn on "advanced reasoning" and retry for it to see its mistake.
roystgnr
2d ago
I didn't take the search engine enshittification discussion a couple months ago seriously at the time, but I think I've just been a "boil the frog slowly" victim.
I just saw a mention of Jerry Doyle online. I remembered making some comment when he died a decade ago, but a quick search couldn't find it. This time, though, I didn't just give up, because I clearly remembered making those comments - my father was in an operating room at death's door at the time, a Metafilter obituary thread was an ironic choice of topic with which to try to distract myself, and ten years later I couldn't tell you what I typed but I still remember where in the hospital waiting room I was sitting while typing.
How bad are things? Google thinks only one thread mentioning Jerry Doyle exists on Metafilter, and it wasn't the obituary thread. Gemini at first thinks an obituary thread existed, but then can't find a URL and swears it must have made a mistake. Bing seems to think nobody on Metafilter has ever talked about Jerry Doyle.
I could swear I remember DuckDuckGo being a search engine at one point, but today it just takes me to a GPT instance that is
not able to retrieve any relevant results for "jerry doyle" site:metafilter.com via my web search. If you paste a Metafilter thread URL (or the exact post title/text you saw), I can help identify which “Jerry Doyle” it refers to and summarize what it says."
This is only helpful in the sense that it's really funny for a search engine to ask me to search the web for it. Any Resemblance To Actual Future Is Purely Coincidental.
But there were winners here: Brave search (3 hits, including the one I wanted) and Kagi search (4 or 5 hits, likewise a success). It should not have felt this amazing to find web search engines in the modern era that actually index the web and then let you search it.
roystgnr
3d ago
Funny bit from Alexander Wales on Tumblr last week:
I do think it would be funny, just once, to have a Freaky Friday movie that was completely derailed by the existence of the supernatural. Like yeah, we're in each other's bodies, that's freaky, but this fundamentally changes our relationship to the world and all other concerns become secondary to finding out what happened, how, and what this means for humanity as a whole.
roystgnr
3d ago
But most results that come with Lean code will probably be just fine.
Most results produced with an honest human who understands what they're doing will be fine. There were a few corner-case bugs in Lean itself discovered recently, and those might not be the last, but they'll at least be rare. And although you can slip in an "just assume X" to get you to your proof of Y, meaning it's then really a proof of X⇒Y, anyone who really understands the problem will understand that "see what you can prove assuming X" is an intermediate step to proving Y, not the final result.
I still have to say "most" here, because the trouble with a proof written in Lean is that it is written in Lean. If you're working on the frontiers of mathematics then you're almost certainly going to have to formalize your own definitions, and a mistake in formalizing a definition is not a mistake a formal system can detect. A verifier will simply verify that you did indeed prove the precise thing you told it you were proving, even if that thing is not precisely the thing that you were writing in English about proving.
But the big catch here is "honest human who understands what they're doing". There will be people who add new axioms (or who are happy when their LLM adds new axioms) "because that made it run" without realizing that's what they've done. There will be jerks or cranks who do this deliberately. IIRC there was a guy on Twitter X a month or two ago ranting about his wonderful new major proof, all formalized in Lean and everything, who apparently didn't pay attention when his use of the sorry placeholder triggered its warnings. And at the frontiers of mathematics, the honest human who understands what they're doing may end up vastly outnumbered by jerks and cranks after the latter feel "empowered" by AI. Even before LLMs there were relatively low-level bits of mathematics famous for breaking amateurs' minds and turning them into cranks. I assume we'll stave off most of them by getting into the habit of checking new Lean code for "hacks" rather than pure technical validity, and that'll work for any proof regarding stuff that's already defined in mathlib, but it still won't stop the cranks altogether.
roystgnr
4d ago
Manhattan. For the "adventurous" version, I like to use both rye and bourbon, add a dash of Angostura bitters, a dash of Cherry Heering, shake with ice and strain into a martini glass with a maraschino cherry or two skewered for a garnish, but the simplest "2 oz any whiskey, 1 oz sweet vermouth, over ice" is already pretty good.
If you like whiskey sours and you make your own sour mix, try reducing the sugar a bit and adding amaretto. Amaretto sours are too weak and too sweet, but if you usee whiskey to fix the one problem and tweak the mix to fix the other then it's a great flavor.
roystgnr
6d ago
at a reasonable pace
I feel like this is the only catch ... but how much of a catch?
Wiki says your 100 units of bioelectric energy will power Speed Enhancement for 150 seconds, which increases your running speed by a max of 80%. 80% of athletic-human running power is going to be something like 250 watts, and 150 seconds of that is about 10 Wh, which means your recharge time at a decent 1kW power outlet is only going to take you under 40 seconds.
Yeah, other assumptions might bump me up by a factor of 2 here, but that still seems fine. Even if we assume that standard electrical power can only be converted into "bioelectric energy" slowly, 10 Wh is small for a phone battery; just keep a supercapacitor or something in your pocket, charge that quickly, and use that to recharge your bioelectric energy slowly.
I think the only thing that makes the mechanic work is if we assume that making "bioelectric energy" is less like recharging a battery and more like synthesizing a fuel, something that requires slow and/or bulky equipment.
I'm wondering if @loper's 50% figure is worldwide, since I'm not reading that for the US, but even the US stats I'm finding show a significant increase: over the last 60 years per-capita meat consumption is up by something like 25% if you include seafood or 20% if you don't, but we've shifted a bit away from increasingly-expensive beef and toward increasingly-affordable poultry. Within each category, could we be eating more expensive cuts overall? While it's possible to export or import more or less of different cuts, to a first-order approximation each cow we consume provides one cow's-worth of each cut of steak. It could be that we're trying to eat more expensive cuts, and thereby disproportionately bidding up prices there, but it looks like the opposite has happened: ground beef has become increasingly expensive this century, doubled in inflation-adjusted price over 25 years.
More options
Context Copy link