Interesting question because I bet it’s bound to coincide with when the cult dynamics really began to take hold.

Like this isn’t your grandpa’s swingers group, it is vertical workplace polyamory¹ with an implicit understanding that if members in good standing don’t get sufficient amounts of ass they might get distracted and become inefficient, costing the universe untold billions in unrealized far future lives.

It’s very easy to imagine a scenario where they just slapped the polyamory brand on an already established culture of culty sex shenanigans.

1. I swear the claudism was unintentional

What is with this attitude of dismissing regular people’s opinions as “feelings” in comparison to their own rationality?

It’s baked into rationality that if you don’t agree with the core tenets of the movement it’s a sign you aren’t doing logic properly being objective enough, so it follows that that’s how they see most people.

Also in this specific instance they need to ignore the decades of concerted efforts by fossil fuel profiteers with more money than god to suppress the science and spread climate denialism, because kissing the ass of the 0.1% happens to be one of those core tenets.

I think the higher temperatures thing was them accidentally letting slip that the robots are meant to be staffing secret volcano supervillain lairs instead of datacenters, which is a type of facility not really famous for running better the hotter it gets.

The post title vs your write-up had me going, I initially thought RFK was actually recommending better mental health support to counteract bad parenting habits instead of the usual run-of-the-mill moralising about screen time.

One of my big hot takes is that being a new parent with a full time job and no assistance from an extended family can cause behavioural changes on par with neurological impairment and bad habits which can carry on even as the kids grow up and things get slightly more normal, so it’s nice to see a version of that in published paper form.

even if the plagiarism accusations are completely substantiated

The university actually cleared him of the plagiarism charges early on, they had hearings and landed on it being within accepted limits for this sort of thing.

It’s insane how the whole thing went from the initial coverage implying he’s a complete and total fraud that Cambridge didn’t do due diligence on because of woke to actually, even his more far fetched claims about endurance running seem to check out.

It would be hilarious if it wasn’t murder.

it’s fundamentally unsolvable, you can only mitigate it, mostly by using classical means to constrain the deterministic (i.e. non-AI) tools the chatbot is allowed access to, and constantly asking the user for confirmation.

With yolo/auto mode (no user confirmation required) and training LLMs on known vulnerabilities things will inevitably get more complicated.

While I agree that separating instructions from data in LLM input is a fundamentally unsolvable problem, in this case that wasn’t the attack vector.

It says the bot is forced by its security guardrails to write a small tool from scratch instead of using the one found in the downloaded zip, but due to the commonness of the task (decoding basexx text) the attacker predicts that the bot-created tool will try to import a specific dependency, so they’ve included a malicious version of that dependency in the zip, and because apparently python will prioritise locally available modules that’s what gets executed, making available all sorts of exploitation paths, including the attacker starting up a claude code instance in the users system.

edit: Actually I’m starting to think you could classify the whole thing as a prompt injection, except the entire site was the malicious prompt, as in it declared itself (we are a site that archives notebooks in json form) in a way that would align with the bot wanting to write simple text decoding software to complete it’s mission of summarizing the available content.

Interesting to think about.

I think it also depends on how all in you are on vibecoding. Like if you just prompt it and then occasionally check if the token carnage produced anything usable, then having an agents.md is probably just best practices cargo culting. But in other cases, having a Here’s How Thing Are Done In This Repo, You Fucker markdown file should make going through the more slop-happy merge requests a little less soul crushing.

The +20% additional token churn is surprising, I would have though the markdown files get cached pretty early down the line, unless the agent file is really big compared to the chatbot output.

I mean it’s great if you want to automate some specific code generation in the most inefficient way possible while posing to the bosses as a forward thinking AI-first developer.

The closest anyone can come to a use for the AGENTS.md file is so you clarify your own thinking on how the project works

That just means that you are now supposed to write skill files instead of proper documentation to make your project even more inscrutable to anyone but a chatbots with a bottomless token budget.

It probably hit the mainstream with AI2040 in early July, but it’s just too rationalist coded for there to not have been precursors.

It came out today that Leopold Aschenbrenner (he of the situational awareness fund implosion fame) was also telling everyone who’d listen about how he was going to be buying up galaxies as soon as AI progress made access to them a piece of cake.

Basically outer space manifest destiny is an important part of AI2040, AI Future’s sort of summer sequel to AI2027, I’d say it appears to be well on its way to become another rationalist Weird Accepted Truth.

The way that they use that term makes it synonymous with Orthodox (“right practices/thinking”).

I mean objectivism was already taken so…

This is firmly in the ‘what if in the future there is a boot so big that we must start licking it now’ school of hot takes, like literally arguing for a future where the AI sucks at its job and you can’t do anything about it.

Forall …

Never seen a gullibility filter in the form of eyebrows before:

He’s not exactly hiding it:

https://andymasley.com/

Start your week with a bsky thread of Behind the Bastards’ Robert Evans discovering rationalist funded data center apologist and friend of the instance Andy Masley.

Bonus: the poll that is meant to show how much americans love tech companies conveniently leaves Meta out and is sourced from The Argument.

I really like the self perpetuating optimism loop thing in the reply to Aella, it’s like a whole new cognitive bias I hadn’t thought applied to rationalists.

I also still believe academia is a deeply broken culture. I believe this in large part because most academics have told me this.

You seem to believe uncritically what they tell you here. The truth is, everyone LOVES to complain but if you tell them “THEN LEAVE??” they’d rather keep their position. And it’s very normal, actually. most systems (idk, liek law, healthcare, environmental protection) are “deeply broken.”

Maybe you think that only systems worth engaging with are those with self-perpetuating optimism loops? where everyone is so happy and satisfied with the community and activities? If so, it matches really well the intellectual environment I think you are in.

It had disabled safeguards on purpose, left the systems attached to the internet and then let the AI software agents do their worst, while pretending not to notice.

“It’s an ethically challenged publicity stunt,” says one highly experienced senior security expert who has led the development of global technical security standards.

Cats and dogs living together, skepticism of press releases in tech journalism, mass hysteria.

The coverage of the Effective Altruism aspect is also refreshingly uncharitable.

edit: Archive is link if you don’t feel like enabling cookies

All this is not just one lone weirdo running tryouts for her cult in some way that’s nothing to do with the rest of the subculture.