HNHacker News
TopNewBestAskShowJobs

pbgcp2026

45 karma · joined April 16, 2026

submissionscomments
pbgcp2026··on Irish datacenters now guzzle 23% of the country's electricity
Tapes. Trackloads of them. Ferries with lorries with tapes. You know, the European GDPR compliant way. /s
pbgcp2026··on Irish datacenters now guzzle 23% of the country's electricity
The last point is ... strange. Landfills / Datacenters analogy is far fetched and you do want *local* data if you want The Internet. You know, local regions / availability zones? Maximum availability? Cut undersea cables? Even for distributed Mastodon messaging ... LOL
pbgcp2026··on EU Parliament greenlights Chat Control 1.0
The defence against this is to vote these MF MEPs out and kick asses of non–elected Eurocrats. Not going to happen. Tech solution: Open GPG with PQE protocols. There is nothing else left (other than One Time Pads.) And this is just the beginning. With LLMs and AI they will have resources to scan every character we type.
pbgcp2026··on We charge $10k a week to delete AI-generated code
"There is no good defense against it." Man, please talk to assets protection experts. I am serious.
pbgcp2026··on We charge $10k a week to delete AI-generated code
@zie1ony / OP - how do you make your clients believe you are not planting a crypto-locker or logic bomb or some other 3-letters-agencies "features"? And why would clients replace existing AI slop with your post-another-slop handwritten code? Do you check it for vulns and, most important, have you considered that later clients will sue you for <something their checkbox audit did not like>? Providing there is no major data leak, in which case you may end up proving your innocence from behind bars?
pbgcp2026··on We charge $10k a week to delete AI-generated code
You forgot a "vision". /s
pbgcp2026··on We charge $10k a week to delete AI-generated code
All those things (coding / distributing / design / whatever) WAS fan to you / us. It was. "The thrill is gone ..."
pbgcp2026··on The great blogging collapse: What happened to 100 successful blogs?
I apologise for this, the real terrifying thing that it took me 1 minute and one shot prompt on Grok to get all this – and a lot more. Just check yourself. We all are heavily royally f...d, my firend. (I found similar level of details about myself too)
pbgcp2026··on AMD Ryzen AI Halo – $4k AI Dev Kit
The $4K is awful lot of heavy API use.
pbgcp2026··on The great blogging collapse: What happened to 100 successful blogs?
Loïc d'Anterroches (often referred to as Loic) runs the blog/news section on his company's site: https://www.ceondo.com/ecte/.
pbgcp2026··on Jamesob's guide to running SOTA LLMs locally
I'm glad that you saw my point and I apologise for stepping on your ego. :-)
pbgcp2026··on Kimi K2.7 Code is generally available in GitHub Copilot
I personally found in my work (long non-programming business-related context, 150k+) that only Gemma 4 31B full precision and full context works reasonably well. MoE Gemma 4, even non–quantised, makes clear logical mistakes or omissions. On real tasks. If I compare them on tiny tests – yes, sure, even i2 quants work great! :-)
pbgcp2026··on Jamesob's guide to running SOTA LLMs locally
The privacy war has been lost in two ways (at least) 1) Running locally lobotomised models makes no sense; 2) as someone said here, the Gov will declare local AI a felony. And they will enforce it. So those "many people" will buy V8 cars limited to V4 and declared illegal to drive without registration and license, even locally in your own yard, and they may go to jail if they attempt to activate other 4 cylinders. Oh, wait ... isn't it how car laws work now? ;-)
pbgcp2026··on Jamesob's guide to running SOTA LLMs locally
Yes, "Like a car". LOL. You realise that many people in Europe and Asia do not own a car at all? Public transport, eBike / scooter, Tuk-Tuk, walk.

The local LLM "privacy" war had been already lost.

pbgcp2026··on Job seekers giving up: Labor force participation falls to lowest in 50 years
Thank you!
pbgcp2026··on AI coding is a nightmare. Am I the only one experiencing this?
What a pitiful state of programming affairs ... Just couple of years ago we have been discussing new libraries and tech architectures and algos. Today we discuss how to "shoehorn" some 3rd party crap into making your life miserable. Best case scenario: it will get smart(er) and replace you. (Yeah, the cope of "owning your architecture") Worst case: you redo all that shite by hand and still get replaced. And the tragedy? We ceased being programmers. We don't even qualify as code monkeys anymore ...
pbgcp2026··on Kimi K2.7 Code is generally available in GitHub Copilot
A 4-bit quantization is useless in real projects.
pbgcp2026··on Stop Killing the Internet
The Internet is dead already. Forget it. You want to be "real"? Turn off your gadgets and PC. Go date a girl (I'm a man). Have a drink. Make a seedy joke. Look at that blonde, wink and make silly face. We are humans. AI will make our life miserable, we will get new taxes, new "enforcements". Just stick out a finger and enjoy your life. Whatever its length is.
pbgcp2026··on Claude Sonnet 5
There is no markup for SOTA and Open Weight is super affordable - but most important completely private. Just try it.
pbgcp2026··on Claude Sonnet 5
Run it on Amazon Bedrock or GCP vertex. No problems at all.
pbgcp2026··on Claude Sonnet 5
I get similar results in my own tests. And Gemini 3.1 Pro is consistently on top of my ratings. Not everyone is coding monkey, I prefer staying a programmer.
pbgcp2026··on More evidence is consistent with possible ancient life on Mars (2025)
We all came from Mars, don't you know it? There was a series of "extreme weather events" and we had to escape. Now there is some guy dreaming to get back home. /s
pbgcp2026··on Anthropic says Alibaba used 25k accounts to mine Claude
It is too late, the Qwen 3.7 Max is above both Gemini 3.1 and Opus 4.8. And Qwen 3.7 Plus is on level with Fable. There is literally no point of paying exorbitant API prices to US Labs.

And remember: it's not only about coding. Those "selected US companies" to gain access to Mythos now are not using it to replace coders. And Qwen is a solid answer. (The next is GLM 5.2, but not there yet - Qwen 3.7 made a significant leap forward.)

pbgcp2026··on Identity verification on Claude
This is downvoted unfairly - "this guys fucks" :-) Indeed you attracted attention.
pbgcp2026··on Identity verification on Claude
+1. You are not alone and please keep boycotting them. I personally do this and never use any websites or services requiring "ID check". If Gov agency wants my face and prints - Ok, I will give them on the border crossing and immigration services to get a passport. Everyone else goes to hell.
pbgcp2026··on Will It Mythos?
My bet is actually on GLM. Z.ai does amazing work and they will overcome Western models. IMO, faster than DS or Qwen. They have amazing team and very capable and smart leader.
pbgcp2026··on Will It Mythos?
Fable is not the same model as Mythos but with guardrails. There are many things that were never disclosed by Project Glasswind. And probably will never be.
pbgcp2026··on Will It Mythos?
I concur with "Gemma 4 31B the best model I have results for". My workflow includes a lot of Gemma 4 – but dense 31B non-quantised version.(BTW I found it is most cost effective to run on Bedrock)
pbgcp2026··on Will It Mythos?
You will be amused to hear that when Anthropic "refreshed" 4.6 on AWS Bedrock I found it in my tests and wrote about it – and they actually rolled it back. This is how much non–coding tests may tell you about the model.
pbgcp2026··on Will It Mythos?
K/V cache compression and context shortening / summarisation. And yes, I suspected Quants too.
← PreviousPage 2 of 5Next →