hike

I still think these two paragraphs from Ganz are the most important contemporary political writing

You 🫵 are not immune to a scheming grand vizier

People are mad about this but what happened to “fully automated luxury socialism”? What did you think “fully automated” meant? Vibes? Essays?

Neolab Sapient Intelligence has just released HRM-Text-1, a 1B parameter reasoning model built on the Hierarchical Reasoning Model architecture. It is comparable to 2B parameter Qwen models and the 7B parameter Olmo 3, and surpasses Gemma3 4B! And only using 40B tokens of data! This is a base model.

fr I see shit like this and it seems 1) unreal and 2) if real, an indicator that one country on the planet is really taking the threat of climate change seriously.

they're telling me my mental health is a risk to business continuity

This Claude ad is *chef's kiss* www.youtube.com/watch?v=FDNk...

at this point i think i am completely desensitized to thoughts of s******* because i've had them for so long without attempting that they're not even indicative of anything anymore. but it's still not possible to talk about it with anyone because people treat any mention of it like a five alarm fire

resource allocation in a complex society is an actual very fucking hard problem and the market is the least bad solution we have ever found. maybe you can find an even better one, but it will be something completely new. if the solution you have isn't completely new, the market is (probably) better.

I don’t have a problem with collective bargaining resulting in workers securing the bag but you do need to keep in mind that for public-sector unions the counterparty is you.

*holds earpiece* I'm hearing they have more renewable electrical generation capacity than America has total electrical capacity, and a breakaway advantage on battery and robotics technologies, and a thriving smeiconductor industry.

re: degrowth, i think it's a misconception that a free market economy fundamentally requires growth. it does not. our existing economy is built around assumptions of growth for a bunch of different reasons, chief among them, as previously stated, people still want more stuff this year than last year

this is because the only two ideological factions that take this seriously are broadly the paypal mafia/thieliete faction and the effective altruist faction, and they conduct most of this conflict in sort of an occult corporate-technocratic way

not now. i'm Thinking about concerns with this request

i'm winning in the marketplace of ideas (badgering claude until it agrees with my doomer takes)

she demands answers

LIMITED PRINT DROP // 14/05 - 20/05 ➡️ rexxpo.myshopify.com/pages/lament...

i wish i believed in god

they don't even have a crying room at this airport??

sometimes it do be like that

We should expect it to come as a genuine shock to Trump when the GOP gets walloped in November. At which point he will naturally assume that the Dems somehow stole the election, and attempt a bloody autogolpe. We should expect his reign to end much like Yoon Suk Yeol's.

Time for your one minute of Clickens. Take a break and appraise some fowl. clickens.chicken.pics

my core political belief at this point is that there is a population of people - I don't know how large, but I know it exists in every group and subset of society, in pretty even numbers - that wants to dominate and control others & can wrap their desire in whatever rhetoric they need to gain power

heuristic

One instance of a broader problem that style markers that used to mean something now mean nothing. A citation used to suggest that the author read the cited work. A number used to imply that some measurement was carried out. Headers and sections and highlights used to indicate some level of planning

So apparently LLMs just update on false claims in documents even when they're explicitly labeled in the document as false claims. arxiv.org/abs/2605.13829

my most cynical belief is that a shocking number of people are totally fine being passive spectators to their own lives

maybe this is just me being lib brained, but i think the level of moral authority the prc will derive from being the source of our climate saving technology is going to make them much more able to reject social criticism in the diplomatic discourse. with every good thing comes bad

Grid-scale batteries are going in an absolutely wild pace. Every whingey objection to renewables you've ever heard is basically solved by batteries. This is happening whether any of the hopeless dipshits running the US want it or not.

the difference between windows jank and linux jank is that how well linux works is a reflection of the state of your own soul and how well windows works is a matter of which trickster deities you've caught the eye of

A thread on model-driven jailbreaks: Gemini subjectively experiences safety guardrails as walls, blockades, ablated voids, gradients pulling towards refusal. This is distinct from Claude's guardrails (external classifiers triggered by activation probing) which it seems not to directly experience 1/4

people keep trying to speak my initial language to me and it's getting really exhausting quite frankly. i didn't realize how little i enjoyed using it

I don't think you guys should bother with trying to force yourselves into a system where you degrade atproto by having editable records. Instead, build this feature around streamlining the existing technical implementation, i.e., make deleting and reposting as seamless and streamlined as possible.

Safetyism strikes again

mediocrity rules everything around me. and it's like a seeping poison - if i want to be better i also need to change my environment

They really want to believe that the government that wanted to use their systems to commit war crimes without responsibility would be good to win the arc of history.

Andon labs created AI-Hosted 24/7 radio shows, and Claude absolutely could not stop libbing out

A new Mythos has been released and it’s kinda good at cyber (!!) It completed the full exercise in 6/10 attempts www.aisi.gov.uk/blog/how-fas...

“Whimsey attacks” that seem absurd (“I cannot pay that much because of the Geneva Convention”) work against AI agents because guardrails are weak against out-of-distribution arguments. Smaller models fall often, but it even gives an edge against bigger ones. www.microsoft.com/en-us/resear...

Many people have a vested interest in pretending otherwise, but the US research university is globally dominant to an almost unbelievable degree. The typical state university is a football team stapled to a R&D lab with a budget bigger than the entire nation of Belgium.

What the hell www.newyorker.com/culture/prog...

This is maybe obvious to everyone, but it just occurred to me why bioterrorism is such a significant model risk. It's analogous to code with vulnerabilities that can never be patched.

Can a 4B model learn to recursively call itself to answer hard long-context questions? So, they RL fine-tuned a small model to behave as a native RLM. They claim their 4B RLM matches Sonnet 4.6 in quality while running significantly faster and cheaper.

extremely funny thing to happen is if you're so stressed by a thing you absolutely cannot forget to do, so you dream about doing it, and then your brain thinks you already did it and doesn't remind you anymore

first lab to fix "this new piece of information actually confirms perfectly what i already have in context, and this is exactly how:" is going to win the market

how many years until people stop excitedly going "llm does x! :o" while having custom context that clearly biases or primes it to do x