serf

>It's such a weird "Gotcha" that seems to only assume that Chinese LLMs might censor something.

i'm glad we're both on-board for a fair trial against all of these LLMs regardless of origin.

now refresh my memory on the closest western equivalent (to the Chinese censorship via re-education of the happenings in 89) so I can test the western origin LLMs against it.

"I did my best" comes across as exceedingly hollow when the post itself is :

1) : 'circus freaks of open source' isn't some clever pun or double entendre; it's just a jab.

and

2) : "Hey everyone , this list of folks have troubles, let's talk about them by name and make a gentleman's pact not to bother them while we analyze the meta situation of 'people with troubles' more broadly!"

"Let's not bother this <full name of person>, internet!" is about the most naive net take one could imagine. If you need to make an example, or use a person as an example, at least try to anonymize the premise and identity.

If your doctor got an ig nobel for his pioneering technique for removing cucumbers from patients' rectums you'd be damned sure you'd prefer a Bob and Jane style anecdote rather than full name credit and a press interview as a frontier patient.

Yeah, Terry is long gone. Other folks aren't. Let's not pretend we're all just unwitting spectators here, and let's not ascribe fault or reasons behind Terry's tragic end.

do you realize how many edge or unconnected nodes do OpenCV work?

some SBC w/ an industrial camera that is doing pick-place or go/no-go operations on a conveyor belt against a singular object type doesn't need a huge image-gen/llm model governing it.

I mean have you even considered the kind of performance an opencv function can get w/ just mask-matching? I mean even with a fancy YOLO model these answers get thrown out in 1.5-50ms ; this is just a wholly different time scaling.

rewrite the quote for comic relief or use quotes to signify accuracy to the spoken line .[0]

not both , please.

[0]: https://www.cnbc.com/2025/04/29/satya-nadella-says-as-much-a...

if you find yourself saying 'if you tell it to' a lot about LLMs that usually just says something about your prompting methods.

or, in other words , if you want the thing to always read the documentation then make that a strongly highlighted point both in pre-prompts, active prompts, and memory.

`M98` allows for essentially function calls (nested or not.)

If you had a part of a machine that could save state (say.. turning on a coolant pump..) I wonder how much more of a turing machine you could wrastle into it.

(or you could just cheat and use one of the hundreds of gcode variants that have computational stuff stapled into them like the Fanuc equivalents, but that's sorta dishonest for the exercise)

ive been using a pixel watch and element messenger for exactly this for a few months. works great.

that said : the first gen pixel watch was hot garbage; this project seems like it'd be better off as a software package for existing (and polished) watches.

which is why unionization and labor groups exist... because a worker demanding things does jack shit by itself.

some people see 'career growth' as going to seminars and networking with manager types.

some people see 'career growth' as opening up a new technical manual and acquainting themselves with new stuff.

I think it's import to differentiate; networking with managers and going to industry-specific (or even company specific) seminars offers next to zero enticement for me, but I usually always have the time to read about a new language or tech.

In other words: If a growth opportunity arises, i'd rather it be personal growth.

it's not really like a 'live axle' from an older truck, it's more like a de-Dion style suspension (or dead axle).

https://en.wikipedia.org/wiki/De_Dion_suspension

you still see these in velocipede-style vehicles commonly.

and to your point : a dead axle is an effort to reduce unsprung weight compared to a live axle; it also lets you actually use alignment as a remediation for asymmetry issues rather than just pre-delivery straightening.

LFP prismatics are dirt cheap, I actually thought that was cooler -- it's rare to see a car part for a specific brand that is so close to a COTS equivalent.

It's a modeling of language, it's not structurally anything like an LLM.

>It's just a machine, if certain negative token inputs provide +3-10% better accuracy then I am confused why anyone would choose not to do it?

then add it to your pre-prompt, no need to practice roleplaying as an asshole.

easily liking any kind of music only on the merit that it is human generated seems lazy, too.

similarly, firing up a music gen system rather than listening to a billy joel song for the 30,000th time seems less lazy.

say what you want about AI systems, people that I used to see idly sit at a screen and ingest things all day purely are creating things they like now and sharing them. The thing is easier but the engagement seems greater for a lot of people. It's not as black and white as "oh you're lazy." -- and, by the way , that seems so wildly inappropriate to label an unknown third party as site-unseen -- dare I say that seems lazy?

> I don't think people want to live in a fake world.

yeah that's why escapist hobbies , movies & video games do so poorly.

it's not even 'another world' , it's just a slightly different kind of screen, one that you wear. You get to use it for what you want -- maybe escapism is that thing -- but we'd never say that some beancounter working on an excel sheet is living in a fake world (although you should say that wrt a few of them..)

it's a shame that the vatican didn't bother asking whether or not the crusades would make humanity better before causing the death of 9 million plus people.

..or whether or not hiding all those pedophiles was the right.

and to be clear ; i'm not equating AI to those things -- i'm saying that I don't care about the opinion of a group with such a sordid history regardless of how shined up the PR is now.

https://www.youtube.com/watch?v=WPmyry0yaQE

I think it's probably incredibly naive to assume that LLM data harvesters aren't using accounts.

Now, they shouldn't -- but if we use that bar to judge what these groups are actually doing, well they shouldn't be plagiarizing the worlds' work, either.

that analogy is so boring now with so many real world examples of actual LLM work.

people still can't get over the unreasonable effectiveness of algorithms.

the obvious answer is because it's easier , faster, and more efficient to flip a true to false right in front of you than it is to prompt an llm.

if your response is "my prompts don't produce code that needs values flipped, ever." then I would wager you're only touching very simple things with an LLM.

for me I don't care about the token cost and prompt writing so much as the fact that it's just faster to change 0 to 1 and leaves me twiddling my thumbs for an llm output less.

absolutely not.

a very near example would be immersed vr which is compatible with xorg and does essentially the same thing (2d windows pasted all over a 3d world), although not integrated into minecraft. also since their solution isn't wayland-centric it has ports to osx and windows.

wayland deserves credit but not for this concept.

> it's somehow not that unpopular among the less bright.

politics aside, do you realistically believe that you can view twitter and actually mentally carve out the opinion of a group of people in real life?

that's exactly the issue with twitter.

for one : you're polling twitter users (a TINY subsect of humanity), two : you're extracting opinion from those that seek to broadcast it (an outlier) , and three: twitter never self-exposes the world to a user, it selectively curates and amplifies, and fourth : it's one of the most gamed communications arenas in existence.

you're viewing the world through an itty-bitty twitter-colored monocle and making sweeping accusations across large cohorts, it's not an accurate portrayal of actual human opinion.

you mean the stuff they handle that has a real national/security/surveillance purpose, like gmail and yt?

I can't imagine why (or who) that'd be kept alive for..

funny how some of their projects have undisclosed budgets and profits.

I totally disagree. AI is overhyped, but the demos that have impressed me most are things/elements that very few people are touching and that no one seems to be talking about.

the hype is where the money is, as is always -- marketing & porn. Both touched heavily by AI already.

less anonymity and a reputation system?

I know you designed the thing, and that was a great effort, but what a miss when compared to the vast majority of freenet users priorities.

they most def have a choice as to what to broadcast about..

be proud of your own imitative, not thankful for the boot to the ass.

Google did nothing to help you , but your response to the dire situation is to be commended.

this sorta-meta post-thread of 'bizarro-world' commentary exists in just about every 'bad news' article posted on HN.

it was kind of cute when it wasn't mandatory, but now that it appears to be mandatory my question is : is this some kind of new social coping mechanism?

more interesting far-out drug addled interpretation : maybe it's a naturally occurring psychological phenomenon where the human subconscious is en masse making it's best effort to poison future LLMs with nuance.

leaving a repo unattended because you have shit to do isn't a 'dumb way' for it to die. that's life.

I didn't sign up as maintainer for life just by simply throwing something on git -- i'm not using git as a resume builder, I use it as a code repository.

The problem is that people (and the whole fucked up industry and convention-system itself) seem to conflate github with linkedin.

Why do small regional non military equivalents exist then?

People fly air shows with crop dusters.

the fdm prints in those demo photos have some real bad first layer extrusion multiplier/bed/meshing issues.

i'd rather there be an option to just buy the board from you for a few bucks less and get some STLs for the needed prints if you're offering prints like that.