TESS

Welcome to earth. There's no magic here except what you bring with you.

~GOOD FOR MAGICAL GIRLS. PRODUCES EXCESS ENRGY GENDER: ʌ ʌ / ⦿ ̮ ̮ ⦿\ ⏝/ | ⏝

@madeofmistak3 when the basilisk was trained on youtube

at the $10 and up patreon level i will not trap you in my "killbox"

@LabGrownNeet @liz_love_lace HOLY SHIT

@liz_love_lace @xsphi EATING OVER THE SINK IS OPTIMAL FOOD LOGISTICS

GIRL WHO'S WORKING TOWARDS THE POST-SCARCITY INTERSTELLAR TECHNO-UTOPIA HUMAN EMPURE BUT DOESN'T BELIEVE DISH WASHERS ARE USEFUL AND ISN'T ENTIRELY SOLD ON VACUUM CLEANERS EITHER

DISH WASHERS ARE GOOD FOR PARTIES/LARGE FAMILIES AND VACUUMS ARE GOOD FOR CARPETS, GRANTED. I'M THANKFUL TO HAVE BOTH. BUT I THINK A ROBOT WITH ARMS THAT CAN SWEEP AND HAND-WASH DISHES WOULD BE A BETTER SOLUTION

WHAT IF BOTH OF THESE THINGS MADE AN UPSETTING SOUND FOR THE DURATION OF THEIR USE?

@SenSanders or analyze people's movements from public cameras and other sources. or use that same metadata collected for ad targeting to *precisely* analyze each persons habits. a whole new world of possibilities is opening up that nobody has any experience with or intuitions about.

@SenSanders a government could, hypothetically, now have LLMs read every single text message of everyone and arrest all "dangerous potential terrorists" (people who criticize the regime). this simply wasn't practical before, or at least took a huge lift.

I spoke to Anthropic’s AI agent Claude about AI collecting massive amounts of personal data and how that information is being used to violate our privacy rights. What an AI agent says about the dangers of AI is shocking and should wake us up.

@phantomfireworx @lossfunk yeah, for the easy ones. I also notice the numbers don't go over 255, which makes it a lot lot simpler assuming the LLM is told it's allowed to make this assumption. even so the harder problems are hard.

@xsphi @lossfunk first pt is absolutely true but at least for brainfuck you could probably do the easy/medium problems by hand since it's just a really god awful turing machine

🚨 Shocking: Frontier LLMs score 85-95% on standard coding benchmarks. We gave them equivalent problems in languages they couldn't have memorized. They collapsed to 0-11%. Presenting EsoLang-Bench. Accepted to the Logical Reasoning and ICBINB workshops at ICLR 2026 🧵

@JarlsberoMan @lossfunk turing completeness means it is theoretically possible to, at the very least, build a transpiler from normal language to esolang. but this can be really hard. if this is the easiest way to solve the problems it's no wonder LLMs can't do it. most programmers can't either.

@JarlsberoMan @lossfunk when you're programming in brainfuck (and presumably the others, though I haven't used all of them), the vast majority of required reasoning is to implement things that are trivial, atomic operations in normal languages. for example "divide X by Y" is a complicated program

@cantoni the ingroup. I think they're cool :)

Feeling agreeable, what's a good thing to agree with for someone just getting into agreeableness?

@lossfunk "All requiring identical reasoning to Python" this is not remotely true. these languages are inherently hard to program in. do you even have human-written existence proofs of reasonably sized programs that solve each problem in each esolang?

2/ Our method: test them on esoteric programming languages. Brainfuck. Befunge-98. Whitespace. Unlambda. Shakespeare. All Turing-complete. All requiring identical reasoning to Python. All with 1,000-100,000x fewer GitHub repos than mainstream languages. Same problems. Radically less training data.

@inerati boooo

@xsphi i’m not roming it

i was forced against my will to upgrade to ios 26 and now the keyboard has round corners but only sometimes depending which app you’re in. also everything feels super big did the icons get bigger am i going insane. also why is it lagging when i type.