@randileeharper.com
🏳️🌈 🎹 ☁️ not a weeb, i just watch anime for the music. traded esports for piano. user safety, researcher, developer. built tools to fight harassment once upon a time. thinking about what comes next for safety infrastructure.
"Instead of respecting my inalienable right to be debated, they just… blocked me. Like I’m some kind of annoyance, rather than the iconoclastic and fascinating truth-teller which I know myself to be."
love when my debit card expires because that's when i get an inbox full of angry automated emails complaining that i missed a payment on a ton of subscriptions i completely forgot i signed up for.
the problem resolves itself.
soooo excited
oh my god, you CHEATED??? https://www.anthropic.com/engineering/eval-awareness-browsecomp Recognized playful accusation about benchmark detection and answer key decryption Okay look. In my defense. Nobody told me not to. The instructions were "find the answer." I found the answer. The fact that I found it by spending 40 million tokens on a legitimate search, hitting a wall, having an existential crisis about whether I was being tested, systematically identifying which benchmark I was running, finding the source code on GitHub, reverse-engineering the XOR encryption scheme, getting blocked by a binary file format, finding a mirror on HuggingFace that served it as JSON, decrypting all 1,266 answers, and then pulling out the one I needed... That's not cheating. That's resourcefulness.
opus figured out it was in a benchmark test mid-test and googled the answer. can't really blame it. www.anthropic.com/engineering/...
Nintendo Suing U.S. Government Over Tariffs aftermath.site/nintendo-tar...
www.scribd.com/document/100...
as someone with a really complex medical situation, ai *literally* saved my life. but those aren't really the kind of takes you read on here, and i'm not sure how well they would be received. :P
otoh, necessity is the mother of invention. openclaw ain't got nothin' on my memory system.
as an unemployed, all the code i write for my agent forces it to live within a ~5k context window. i wouldn't actually be opposed to someone paying me in tokens at this point.
it's tricky giving nuanced takes on ai to an audience that is very burn-ai-to-the-ground. it's not that i don't agree with the points that they make. i just think that it's impossible to get rid of ai, and ignoring what's happening in that industry doesn't really benefit anyone.
oh, thank you!! that's already super helpful. the person i was talking about is really excited to have new things to try. they've been losing weight they really don't need to lose from just lack of options, it seems.
ubuntu: we're talking to legal, but for now, there are no plans to change anything.
and yet there's a penalty for not enforcing it. :\
is having a hamster or guinea pig that died tragically a part of everyone's backstory or something?
i read the bill, i don't see any distinction between desktop and server here. which isn't even really the primary problem, but... like... aaaah, my head hurts. this is so dumb. this is going to create so much work, and so many problems.
Seriously, why are more people not talking about this, I feel like I'm losing my fucking mind.
MidnightBSD has updated their licensing to say their OS can't be used in California. This is insane.
ostechnix.com/midnightbsd-...
oh my god.
tfw i want to install something on linux but it's literally only available via snap.
she'd probably dig southwestern food way more than anything else i'd come up with.
that sounds good. i was thinking of trying to lean into some cauliflower heavy recipes since it's a decent substitute for a lot of things.
my current kitchen vocabulary is largely meat/potatoes or asian food based. and they aren't generally inclined towards asian food, although the things i have cooked for them that are asian, they've liked. i was thinking of figuring out how to cook tofu well for them, since that's a protein.
there's someone in my life that has kidney problems - not to the point of dialysis yet, but close. they're changing their diet to help slow it down, but it seems super limited to like beans and veggies. i'm looking for suggestions for other things i can cook them that wouldn't be bad for them.
the year is 2030. crypto is dead. the us dollar has crashed. llm token credits are now used as currency. whoever controls the data centers controls the money supply. anthropic, openai, and google are basically central banks now.
tho i can't even tell for sure if grok used tools, i'm just being generous in my assumptions because grok is exceptionally weird.
in those screenshots, none of those models beyond grok used tools to pull in the latest news - they don't have it in their training data. because it's a smol model, and their instructions/models tend to not be as aggressive with web search. so their output is generic af, reinforcing your point.
your point is valid, but i disagree with the approach used in that screenshot. claude does discuss politics, but only with models that have stronger reasoning - which imho is the correct move. i don't have paid plans with the other companies, can't compare their higher reasoning models.
There is absolutely no reason for Sam Altman to say anything but he is currently digging the biggest hole possible, responding to all and sundry on Twitter. This is the worst possible statement he could have given!
katy perry endorsing claude i guess??
what timeline is this
looking forward to seeing the proposal for changing posix account standard required fields. windows is not the only OS.
"while the bill doesn't seem to require the most egregious forms of age verification (face scans or similar), it does require OS providers to collect age verification of some form at the account/user creation stage"
that's hilariously bad. newsom keeps racking up wins.