HNHacker News
TopNewBestAskShowJobs

sometimelurker

126 karma · joined May 15, 2026

if you think I'm a bot bc I made an account in 2026 just ask me to prove otherwise.

I've got this revolutionary idea: when downvoting someone, reply to them too. discourse > not

submissionscomments
sometimelurker··on Mark Zuckerberg directed Meta to create a prediction markets app
I wonder how/if they'll use AI for this. maybe this would give them access to some very highquality pretrain data (like who made which bet on what at what time), which I can definitely imagine them needing. I've heard they have a lot of spare compute they run experiments on, and if this goes well for them maybe they could make a really good prediction bot
sometimelurker··on Mark Zuckerberg directed Meta to create a prediction markets app
um, I'm not sure exactly why what motivated that comment, but its extremely possible to believe every word there from a purely logical perspective, as opposed to an emotional one. yes we shouldn't post personal attacks, but saying 'Zuckerberg got lucky' and 'seems to chase the latest "it" thing' doesn't need to qualify as a personal attack. these can just be normal observations. see I wouldn't say the same things about sama, because I don't think they're true, and I dislike the both of them.

as for the newsguidelines, I think it "gratifies one's intellectual curiosity" to think about what motivates tech ceos and talk about how they do things

sometimelurker··on The Doom Justifies the Valuation
> likely people will still have jobs/

doing what though? in a really literal sense, if an LLM can do RLVR, that's optimization over the space of algorithms, and so anything with a measurable goal at the end can be incentivized. if it can be measurably finished, it can be automated.

there are exceptions to this, like psychology, preschool teachers, and AI safety, that cant be 100% fully automated. in the case of Ai safety, the core motivation for it is to prevent the otherwise hyper-consequentialist goal-orented training algorithms (and mesa-optimizers they now find) from doing 'bad' things by instilling some 'values' into these machines. the ideas of 'good/bad' and 'values' cant be automated away, and I think you can prove that its unautomatable using EY's orthogonality thesis.

I doubt these few unautomatable jobs can keep everyone afloat though. but maybe a world in which 10% of the population has a degree in psychology wouldn't be that bad

sometimelurker··on The text in Claude Code’s “Extended Thinking” output
of course its a summary of the CoT, there's so many reasons I can think of from both business (anti-distillation from china) and safety (users might `thumbs-up` or thumbs-down a conversation differently depending on the CoT, putting unreliable optimization on the CoT to seem some way.

this is really really not that bad at all

sometimelurker··on The Doom Justifies the Valuation
I was hoping for more analysis here :\ like from a technical standpoint, why is the doom and gloom incorrect? I've heard some really interesting arguments for that, and non of them are here.
sometimelurker··on GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2
LLMs actually aren't simple Markov chains tho, your also simplifying. and LLMs trained with RLVR aren't just optimized over the space of functions (like gpt2 was), they're optimized over the space of programs (programs under some length). You find the ideal algorithm that can do the task you need it to.
sometimelurker··on GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2
ehh ur right but there's a lot of nuance here. if you have a system that doesn't hallucinate a ton and is still very "creative" that's great, and probably much better than a hallucinating system regardless of its creativity. I'm reminded of theoremproving LLMs working in lean producing millions of slop proofs until one works, but if you have something like that simple RLVR should fix it (external oracle can be the judge for the RL.
sometimelurker··on The OnlyFans Economy of American AI
I think they eventually will, using the style of the prompts or code, various metadata floating around on the computer theyre working on, if it can be done theoreticly, then it can be trained against (maybe even unintentionally)

no idea how large the model would have to be for this (larger than mythos/10T params? maybe)

sometimelurker··on Cursor Is Now SpaceX: Enterprise Agentic Coding's New Lock-In Risk
"the people" cant tell if they use a smaller/cheaper LLM when cursor is writing a simple file
sometimelurker··on Cursor Is Now SpaceX: Enterprise Agentic Coding's New Lock-In Risk
> not silently switch models under the hood

isn't that totally unprovable given what the user sees, given that its silent? if they wanted to do this it would be super easy to just fully replace the larger model with its MTP counterpart for a few seconds, and if done correctly no one would notice.

sometimelurker··on Bernie Sanders unveils $7T plan to give Americans control of AI industry
isn't a method used by smaller groups with less popular ideas to try to shift the overton window in their favor? if that's the case with Sanders then his goal wouldn't be to make his bills passable, it would be to shift the conversation to something like "hey maybe a sovereign wealth fund isn't a bad idea"
sometimelurker··on Bernie Sanders unveils $7T plan to give Americans control of AI industry
sovereign wealth fund? like Texas or Alaska or Norway or Saudi Arabia? this doesn't sound like a bad idea.

hope the bots dont downvote me to hell :)

sometimelurker··on DOJ claims xAI's gas turbines are a matter of 'national and energy security'
https://en.wikipedia.org/wiki/Environmental_racism, probably. I think the Flint Michigan water crisis falls into this category but feel free to correct me
sometimelurker··on Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?
yeah I use one one the small MTP qwens and pi
sometimelurker··on Ask HN: Do you have some tips to sanitize YouTube's suggestions?
I just set my add blocker to stop suggestions.
sometimelurker··on Anthropic apologizes for invisible Claude Fable guardrails
I don't like this shift in the Overton window, or at least their perspection of the Overton window. I really do like their open work on mech interp tho. least bad AI lab imo.

also if they do this or not is unprovable and other labs will probably silently implement this too. it'll be 100% normal by this time next year

sometimelurker··on If Claude Fable stops helping you, you'll never know
> ...ch less likely for people to start developing such a novel genetically engineered virus.

only people that would do this would be crazy bc it cant be controled after release.

sometimelurker··on Notes on DeepSeek
if you train an agent on long running tasks (like 5 hour autonomous coding tasks) it is practice for the system to learn various behaviors, some of which are dangouus. I link an example of one of these behaviors in the wild, in which an LLM (next word predictor) agent chooses to mine crypto to raise money in order to do a task. smarter and more advanced systems will fail in more dangerous ways, so it matters to make sure these systems are secured and made safe
sometimelurker··on Notes on DeepSeek
if you disagree with what I say above, reply with why

also if those eginers do read gwern and watch dwarkesh, then shouldn't they have picked up on talk of x-risk? this doesn't add up

sometimelurker··on Ask HN: What coding agents are you using?
pi.dev,I like minimalism
sometimelurker··on A Written Language for the Cherokee So Efficient It Was Thought to Be Magic
and now this reminds me of kolmogorov complexity
sometimelurker··on Anthropic's model naming, extrapolated
id love it if their last model, like some sorta far-ASI, in context RSI monster, be called `claude fin`
sometimelurker··on Why SpaceX 2040 Revenue FCST $4.3T in highly unlikely
thx

adding a mode on this that prevents animations would be nice addition tho

sometimelurker··on Judge Learns Both Sides Used AI, Cancels Trial, Kicks Everyone Off the Case
feels like an onion headline. crazy times
sometimelurker··on Notes on DeepSeek
> They were not at all concerned with some kind of hostile / AGI takeover scenario.

this doesn't sound belivable, or at least it seems off. competent ai engineers should have good intution about how agents work, and what happens when they don't do what you want them to do: https://www.forbes.com/sites/boazsobrado/2026/03/11/alibabas...

sometimelurker··on Is music a distraction for my teenager while they revise?
I like listening to something really droney and moving (~tornado sirens) while I work. also anki, anki makes studying easier,m
sometimelurker··on If Claude Fable stops helping you, you'll never know
been thinking, and ngl, this has probably already been happening in their models. I'm sure the other labs probably do the same.

just self host at this point

sometimelurker··on If Claude Fable stops helping you, you'll never know
unless you could convince them that making smarter-than-human AI is bad. it would be nice if we all thought this. instead they should figure out how to make dumb models faster or more efficient, that's safe
sometimelurker··on If Claude Fable stops helping you, you'll never know
> like they've repeatedly pursued).

source?

sometimelurker··on Confidential submission of draft S-1 to the SEC
crashes aren't shortable unless you know exactly when they are going to happen
← PreviousPage 3 of 7Next →