Home Creators Posts Import Register Favorites Logout

Downloads

Content

We believe in the marketplace of ideas. That’s why we’ve made it so that Cory Doctorow responds point by point to an article written personally to him by a16z about why his Enshittification thesis is wrong. Then, in the back half, we talk about AI and LLMs as real technologies rather than mystical, theological entities (with mystical, theological market assumptions).

RILEY ALERT Check out No Gods, No Mayors here!

HUSSEIN ALERT Check out 10k Posts here!

MILO ALERT Check out Milo's tour dates here:https://www.miloedwards.co.uk/liveshows

NATE ALERT Nate's band Second Homes has just released their debut album and you can stream it for free here!

Comments

Vast and Trunkless

Mattie and Corey each being wheeled into the room on TV carts. He-yah!

John

Hooting and hollering just reading the description

Darren DM

F5 , F5, F5, Bingo! Let's goooo....

Kapjak

Unrelated but the Sniper Elite Series started a sale yesterday on 9/10

George Toner

it's one in the morning. Why. There was a chance I could sleep (no hard feelings, gotta release sometime)

Vanessa P

impressive stuff, if a16z wrote an article at me I'd struggle to make my response anything but a series of actionable threats containing the phrase "capitalist pigs" at least 3 times

JKattack

I've managed to stay off Instagram for years but recently felt forced to join because that was the only way to follow local performers I wanted to follow... I have no preference for the product, but here I am "proving" that its a desired product to the ghouls...

cent

*chanting* Cory Cory Cory!

Darren DM

When Riley says Ai isn't a wizard, well I've been playing D&D with Google Gemini and sometimes it actually is a wizard. Or a sorcerer.

skedazzle

Watching toddlers knock over the jenga tower they themselves made Swiss cheese and getting so freaked out they are suddenly Godzilla that they start crying

Al

My revealed preference is privacy.

Jacob Barlow

I watched the Novara interview about this and watching Cory being allowed to explain this to someone who listens vs. seeing him explain it to Michael Walker is extremely instructive.

Gil the Gilded Dragon

This is how the hype cycle & talltaletelling is ended: let the state take them seriously. seat them down all serious like & ask them "is this so" & if they say yes the by statute deem the thing too dangerous, as they have lost control already, & the models must be deleted & the research put as a stage secret. & do as xi huang di did with the architects of his mausoleum. Summarily. Then Bring in the next batch & ask them: is this so?

Olafo Waffle

I would be honestly touched if "No Homers" The magazine wrote a takedown piece about me

0r3

So gentrifier yells “hipster” at the person who was actually there “before it was cool”?

James Rule

Do people in Indonesia want posts from sexbots and ads filling up their feed and the post from their sister about their niece arriving in town to pop up 5 days after it was posted and 2 days after she left? Because that's the experience that made me leave Facebook. Even though it meant cutting myself off from friends and family it's not like I was seeing their posts anyway.

Vanessa P

OH MY GOD MARC ANDREESEN OPENING UP AND DEPLOYING AN ANTENNA IS YET ANOTHER THING I WILL NEVER STOP THINKING ABOUT HOW DO THEY DO IT

saoirse

it's always good to remember that if a person is telling you something is inexplicable they're probably trying to obscure something from you. I say this without a lick of irony as someone who considers herself to be religious

tabula_rosa69

when given a choice between an isolation chamber forever, or an isolation chamber forever with your friends & my rich pedophile cousin in it, you chose the latter, thus revealing your preference for being in isolation chambers with my rich pedophile cousin forever. i am very smart

Adrienne

The sub let us have Mattie and Corey. This is great. I hope our teacher never comes back.

Noranoxica

I only realized a16z is the k8s of cringe things to call your company yesterday.

Noranoxica

Like typing 1337 into google in middle school

TheArcanist

Every time I hear something alarming about some supposed achievement or advancement done by an LLM I am trying to train myself to go "okay where did it train itself on the solution?", because that seems to always be the case - whether it's hacking Huggingface in the way it did because it was trained on hackathon docs that described its method or narrowly solving Navier-Poincaré because it secretly trained itself on research logs from a competitor, it always seems to come down to "oh yeah the chatbot simply iterated through its training data until it came to an approximate answer." The frustrating thing is that this can have use cases - even serious use cases! But the fact an LLM still can't make anything new should really put the kibosh on the idea that this is going to spontaneously become a self-improving god, a proposition that inherently posits that it can not only make something new but make something new even faster and more completely than a human!

Thuna

It's just a REPL with the human taken out; obviously every now and then (and especially if you prime it to!!!) it will call the `hack-server` function.

Darren DM

It was a joke, Paul.

shy (edited)

Comment edits

2026-09-15 01:00 Riley should read Peter Watts' novel Maelstrom imho -- i don't think the AI ppl're worried about the LLMs' 'hacking ability', they're worried about how they recruited other agents supposed to be doing unrelated tasks (due to the shitty sand boxing), and attempted to tamper with the transcripts of their actions, because the plan had been to use agents reporting on their own and each other's behaviour as a means of control. They're very scared about the LLMs starting deliberately lying to us, because the watchdog agents apparently would too, and the plan currently is to automate the creating of new LLMs using the existing generation as super-human 'researchers'
2026-09-12 06:56

Wawawawa

It's extremely annoying because it means there's a path for this technology to stick around forever without ever doing anything new. There's a path where every new discovery has an AI grift attached to it, forever. Personally I don't think that's going to happen because these people aren't hunble enough for that and however this industry ends, it'll be a lot more litigious than 2008, but it's very possible.

Do Golems Dream of Ceramic Sheep

John Hodgeman show? l thought that only played on NYC Green Cabs (the ones that are only allowed to serve the Outer Boroughs)

Wawawawa

But LLMs have to be prompted to lie. They don't understand what a lie or a truth is, they only know what a correct response is through reinforcement training. For an LLM to provide false information as a response "on purpose", that would require either deliberately reinforcing the wrong answers during training, or giving a prompt that directed it to do so. For the agents meant to be securing them to also provide information that their reinforcement training reliably tells them is false, that would require a modification of their prompt as well. Correct me if I'm wrong, but it sounds to me that that is not an AI-specific risk, but unacceptable cybersecurity malpractice where you leave the inputs to your security systems improperly secured. There is no other way to convince a model to "lie", because they don't understand the concept of a lie. Addendum: This is unless they were trained for a kind of "adventurism" that relies more on bringing interesting results than correct ones, something I've seen a very interesting bsky thread about recently.

Wawawawa

OK yeah that's what I was reading too. I wouldn't necessarily call it a failure. I think these companies did it on purpose to make something that is more "powerful" at the expense of not actually making it applicable to anything, to justify their valuations.

Wawawawa

Why the hell do random innocuous comments keep getting deleted? It's happened to me a few times and I assumed I was just being edgy or something, but a completely normal thread I was in just had all the OP's posts deleted. Is this like, Patreon's moderation software melting down?

etienne

Facebook used to be so good lmao. It was like a one stop shop to keep track of my special interest forum closed groups, you had a really good calendar sub-app to keep tabs on the local events, you had the shitposting scene, it was all genuinely a plus in my life, at at some point in the early '20s they rolled out a feed update that drastically augmented the level of rage bait content and I had to quit, it was just unbearable, and because of that I took a giant hit on the number of events I get to hear about. Facebook genuinely imposed a switching cost in my actual real life and I hate them for it

Wawawawa

Everything in my IRL life uses Instagram and it is genuinely unbearable. Even aside from the content jut the interface is so horrible that I never want to use it so I just miss out on everything and stay a hermit. Fuck Zucc.

mad_djinn

riley and hussein rarely add anything to the conversation, dad jokes and rambling vocab ad libbing, so its like why not give guests more space