claude's been getting into xianxia lately

nothing feels as good as blocking someone who deserves it

testing my new standard.site integration btao.org/posts/2025-1...

the scared benchmark. how much does the model scare tao

Had a friend who was a phenomenologist. One day on a job site someone forgot to epiphenomenalize the qualia; when my friend went to wire them into the foundational ground of Dasein he got hit by thousand kilochalmers, right to the pineal. Fried his basis of moral patienthood, he’s on disability now

Productivity doesn't matter if you're working on the wrong thing.

they should make a Loom that is good

......... btao.org/guestbook/

i am such a city slicker

this rocks

The 7 golden rules of not being a dick on the internet 1. No quote dunking. If you disagree with someone, reply to them directly. Broadcasting their post to your audience with a snide caption turns a disagreement into a pile-on. Which is the point, I know. But we're better than that. ...right?

when they say "intelligence too cheap to meter"... we all know that it's definitely going to be metered, right?

looks like sentry is selling codecov to some AI-SDLC company: about.codecov.io/blog/a-new-c... hopefully good news; i found codecov to be extremely buggy & poor quality the last time i tried it. sentry did not seem to be prioritizing it

while we're sharing github-stats side projects... i made this a while back; it shows how many of your github commits were co-authored with AI technically-your-name-is-on-it.btao.org/taobojlen

seems fair cronotype.vercel.app/taobojlen?hi...

What I’m hearing: Instagram’s Trust and Safery org absolutely gutted the last few weeks. ~60% of the org gone - between layoffs and forced reassignments to data labelling. All while “AI maxxing” pushed a bunch of bugs to prod. And hence why today’s massive Instagram account takeover happening.

*looks pleadingly over towards the US from across the atlantic*

i love having bad taste in coffee; it's so nice to be able to enjoy bad instant coffee just as much as idk a fancy cafe coffee

megabonk is very fun

someone should tell tim tebow that we already solved this, 1 like = 1 respect. everyone knows this

this is quite a good website about antibiotic resistance seriouslyresistant.com

this reminded me of the time dan dennett visited my university, and my friend in the philosophy society worked with him to make & sell a hat that david hume also wore web.archive.org/web/20150604...

THIS is what kierkegaard meant????

i have been playing the new ZA/UM game and it's good, but man the ocassional hints of AI writing really kills it for me and i can't even tell if this is pure delusion/paranoia on my part

all of these are real but i wish amnesty had framed it differently... these risks are just as present in the pre-LLM status quo. these are primarily T&S challenges and not particularly LLM challenges, IMO!

opus 4.8 coming in with the typos ("sibding")

Interesting new ai age of "you can have anything you want, instantly at your fingertips, but you have to know what to ask for"

i did not make it, i want to try opus 4.8... anthropic gets another month's existing-in-the-new-world fee

ok i know i am late to the conversation about jqwik, and i think it definitely counts as malware, but fwiw the bot who reported the issue was also incredibly annoying. it's rude to have overly verbose LLMs demand maintainers' attention github.com/jqwik-team/j...

my kingdom, my kingdom for progressive social movements with an orientation towards *winning*

GTA fans have been on the edge of their seats at the rumours a new video is about to drop… Well the wait is over. We present: the public launch of the Rockstar IWGB Game Workers Union. Our platform, to tell our story. youtu.be/iLOq_6rOpoM

opus 4.8 is out: www.anthropic.com/news/claude-... looks like an overall good incremental improvement. the most interesting aspect to me is the increased honesty / less overconfidence

i feel like people crowing about how badly a small cheap model miscounts letters in a word and then writing off the entirety of ai capabilities is as if you showed a human this picture and then wrote them off as completely useless because they said one line was longer than the other

this is very cool news for the trustandsafetyheads out there

i wrote about this recently btao.org/posts/2026-0...

starting to block anyone who posts screenshots of posts to dunk on them. i already know that people are wrong online - why do we need to signal boost it? just move on lol

microblogging is a cursed medium

today i had a sunny date in the park, met some dogs, and now i am drinking rooibos tea + playing the new za/um game. fuck everyone who said i'd never make it

until recently i volunteered at a second-hand bookshop. the majority of donations did not sell on the shop floor, and we sold them on to companies like Better World Books and World of Books. those companies recycle about 80% of the books they buy!

I just derived that I’m right and you’re wrong from first principles. It’s over for you.

this is a little intimidating but i am excited to lean in fully to more open models, for a while at least

perhaps the same goes for embodiment

Pope on AI/Moderation: “The digital economy’s functioning relies on the silent work of millions of people engaged in essential yet largely unseen activities, such as data labeling, model training & content moderation, often involving disturbing material.” cc @julietshen.bsky.social @aaron.bsky.team

sunsetposting

@ohcraphey.app hello! are there any plans to add configurable time thresholds? e.g. in a work profile, i don’t need to keep up with everyone every 7 days — but with close friends i might!

why do people care about the kardashev scale

agreed; who cares if it was GPT-6 or o3 or opus 4.5 or mythos or whatever that first Got Good at maths/cybersecurity/science? the important parts are that a) models are improving exponentially and b) we can now throw loads of thinking at previously thinking-constrained problems, for v little money

Half the internet is people optimizing routines they don’t like for a life they’ve never once paused to confirm they want. The dashboard is gorgeous though.