HNHacker News
TopNewBestAskShowJobs

Xalutiono

15 karma · joined July 20, 2026

submissionscomments
Xalutiono··on Intel Starts Shipping High-NA EUV Silicon
Europe has ARM, Infinion, Siemens and co.

And Trumpf and no Trump so

Xalutiono··on Nobody knows what a used GPU cluster is worth
For a long time it was assumed it wouldn't wear out and tbh if you look how long it takes, how rare it is, its not a real issue...

besides what intel did with Raptor Lake. This series had massive issues with oxidiation. Was the first time ever i became aware of this issue on scale.

Nontheless there are papers out there that silicon can degenerate and does.

Xalutiono··on Nobody knows what a used GPU cluster is worth
Where is this exponential thing coming from?

Why would hardware failure raise exponential with linear scale? Doesn't make sense

Xalutiono··on GigaToken: ~1000x faster Language model tokenization
AI now consumes TWh. This is more energy than a lot of countries on the whole planet.

If this increases efficiency, it has real impact. 0.1% is A LOT at scale.

Xalutiono··on GigaToken: ~1000x faster Language model tokenization
Whats the prefered LLM runtime to use? vLLM?

Tips & Tricks on parameters/settings?

What happens at peak? Do people have to wait now? Increase of latency?

Xalutiono··on OpenAI and Hugging Face address security incident during model evaluation
No its USA / Trump
Xalutiono··on Xiaomi-Robotics-1
Of course I don't. But its a comment so its my personal current view.

Right now we have massive investment into AI and Compute. So massive, its basically unseen. We are adding more compute than ever before.

This compute can do work and it can do now very arbitrary work. Insurance case work, software development, image generation etc. Everything it can't do today, it will be able to do tomorrow.

Now we as humans are getting steamrolled already. I have a handful of work collegues which are worse than an LLM and I would prefer to have them gone just because its easier for me to explain to an LLM what I want/need instead of doing the same thing with these collegues.

But AI is not stopping, every single day we use AI and we give AI hints like a thumbs up or down or complain to the AI that this is not what I wanted, the AI companies get a signal they can train for. We already today have build the biggest global Reinforcement Loop every. We share it with a handful of companies, but every company gets enough RL signals.

If that AI model has learned something, a billion people can use it. So the investment in teaching an AI model a new trick, is a 1000x cheaper than teaching this trick every human.

You might say "okay but i don't want to use AI" but then someone else comes along and does it.

And AI tokens are not cheap. They are expensive and now work has a money value. If I'm a boss and the AI is 10x faster and 10x chceaper than a human, its very easy what i will do: I will reduce my workforce to optimize for AI usage.

While all of the Soft AI stuff is happening, the robotics are leapfrogging progress. Automatisation in self driving trucks also progresses.

We are now in a timeline of full automatisation.

And my personal assumption is, that this hasn't even really started yet because we are 5-15 years away from the real AI AGI progress.

I don't think people will be able to compete with Google, Apple and co. They already have jumped on AI. Google is market leader. And these companies do have the money for a full AI transition.

Besides that, Google has physical infrastructure around the globe, they have their own undersee internet cables.

But for sure it will disrupt software and plenty of companies one way or the other. And if these companes and we as humans are not acting upon fast, we are gone.

Industry is slow and even if disruption means, that the market will still be there but will look very different, disruption hurts a lot for an economy.

Xalutiono··on 1-Bit LLM in the Browser
IT is not black and white and just because you are able to bring up a random example doesn't mean you are right.

These models are build to be small, we have a lot of diverse tasks which do not need fable or whatever because if you are a company and you have to route 100.000 requests per day, costs matter.

And these small models, if they are not good enough for my specific use case, i can easily finetune (which i have btw.).

You can also easily solve your example by having a second stage. If the customer is unhappy about the first thing or an employee flags it as 'wrong', you can either send it automatically to a more expensive model or the employee fixes it and you got automatic training data.

Xalutiono··on Xiaomi-Robotics-1
At some point it will be cheaper to give everyone bread instead of paying for your private military.

But yeah lets see what going to happen.

Xalutiono··on I burned all my tokens researching how to save tokens
You can reduce hallucinations with rules and guardrailes for sure.

And Anthropic clearly shows that we can reduce hallucination in models. Progress is made which you can see in benchmarks.

And at the end of the day, a normal human being doesn't 'hallucinate' but 'missunderstand' things or just doesn't know what they are missing.

A LLM doesn't has to be perfect, it just has to be cheaper with the min. same quality as a human.

So why are you screaming into the void? Because LLM leave deterministic world? A lot of people don't even life in that world anyway.

Xalutiono··on I burned all my tokens researching how to save tokens
My companie spends a few million per month in token.

For me personally, the only reason why i don't use it 24/7 is cost.

Token costs, energy cost, usage limit due to token/subscription cost.

Xalutiono··on 1-Bit LLM in the Browser
Thats absolutly not the point of these models.

They are here to route things, do tool calls, ask experts.

Xalutiono··on Xiaomi-Robotics-1
Robots will take jobs from massive amount of low income people.

AI will take jobs from a massive amount of avg and high income people.

The disruption ahead is getting pillowed right now due to physical limitations of progress like amount of energy, building DC, doing research, training models (its still takes 1-3 month).

Also a massive buffer will be the boomer going into retirement.

Depending on the speed of all of this, we might get a lot poorer before we get abundance everyerwer

Xalutiono··on US gas prices hit an average of $4 a gallon again
Its China. They slashed oil imports massivly. They use their reserves (as a lot of other countries also do) to help asia.
← PreviousPage 2 of 2