Hacker News: vorticalbox

New comment by vorticalbox in "GLM 5.2 and the coming AI margin collapse"

vorticalbox — Tue, 07 Jul 2026 15:52:19 +0000

They are offloading thinking to the AI too.

It then has to look over everything see how it connects together and then decide the best way to do something.

Giving it small and very focused plans when you already understand the system gets it done fast and cheap.

New comment by vorticalbox in "Alibaba to ban Claude Code in workplace over alleged backdoor risks, source says"

vorticalbox — Fri, 03 Jul 2026 10:20:00 +0000

i can see why they want to stop it but 1. you have to pay for the "attack" 2. these AI companies trained on copyrighted content without permission or attribution to anyone who's data was used to train.

New comment by vorticalbox in "CursorBench 3.1"

vorticalbox — Thu, 02 Jul 2026 11:15:52 +0000

this is my finding too, i have moved to it fully for most of the plan/coding.

for most tasks is capable and very cheap, for a days worth of tasks is costing about $10

New comment by vorticalbox in "The KIDS Act would require age checks to get online"

vorticalbox — Sun, 28 Jun 2026 19:45:33 +0000

Isn’t this already a thing because it requires an adult and a payment method to get a connection to your house.

I’ve already stated who I am when I paid.

New comment by vorticalbox in "U.S. government will decide who gets to use GPT-5.6"

vorticalbox — Fri, 26 Jun 2026 19:31:34 +0000

You want to compact early though as sending the whole chat you will end up with a lot of tokens not in the cache which 1. Costs way more and 2. Will slow the request down as it has to process it all.

New comment by vorticalbox in "Reid Hoffman says SpaceX 'not an AI company', xAI 'complete train wreck'"

vorticalbox — Wed, 24 Jun 2026 13:34:33 +0000

i jump about a lot, for coding gemini and grok are definitely not as strong as gpt 5.5/opus/sonnet/composer.

composer 2.5 is actually very good and use it for a good chunk of tasks.

New comment by vorticalbox in "The text in Claude Code’s “Extended Thinking” output"

vorticalbox — Mon, 22 Jun 2026 16:32:46 +0000

There are actually fine tunes of qwen on opus “thinking” tokens that teach it to think like opus does.

https://huggingface.co/Jackrong/Qwen3.5-27B-Claude-4.6-Opus-...

New comment by vorticalbox in "GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2"

vorticalbox — Sat, 20 Jun 2026 12:19:55 +0000

even if they did it it wouldn't be of much use because correct or not the output was the likely output 100% of the time.

New comment by vorticalbox in "Norway imposes near ban on AI in elementary school"

vorticalbox — Fri, 19 Jun 2026 21:38:00 +0000

Well one can no longer search for information in the big search engines without it just giving you the answer.

This ruins “search and topic and write about it”

New comment by vorticalbox in "Ice water drowning survival of young patient (2025)"

vorticalbox — Fri, 19 Jun 2026 17:26:36 +0000

> This happened after he received CPR for several hours at a cabin on the Hardangervidda plateau before rescue arrived.

This means although his heart wasn’t technically beating, he did have blood being circulated via cpr.

When I read the title I assumed he was alone before rescuing.

New comment by vorticalbox in "Local Qwen isn't a worse Opus, it's a different tool"

vorticalbox — Thu, 18 Jun 2026 08:12:46 +0000

I find opus for planning and sonnet for coding but codex for code review.

New comment by vorticalbox in "Lore – Open source version control system designed for scalability"

vorticalbox — Wed, 17 Jun 2026 21:29:22 +0000

security by only obscurity is bad. Having both is better.

For example say I have a hollowed out wall that is hidden behind a painting.

Just putting my money in the hole is bad once it’s found it’s gone but if I put my money in a safe in the hole. Well now you need to find it and break the safe and a hidden safe is objectively better than just having a safe on the floor because you need to find it first.

New comment by vorticalbox in "GLM-5.2 is the new leading open weights model on Artificial Analysis"

vorticalbox — Wed, 17 Jun 2026 11:00:32 +0000

This is a problem I find with opus is will spend so long thinking then going “but wait what if”

To point where I stop it and simple tell it to “start writing code you can work it out as you go along”

Seems writers block also effects LLM

New comment by vorticalbox in "SpaceX to buy Cursor for $60B"

vorticalbox — Tue, 16 Jun 2026 22:40:47 +0000

Yeah, is that a bad thing? They did some really interesting stuff when fine tuning.

https://cursor.com/blog/composer-2-5#targeted-rl-with-textua...

New comment by vorticalbox in "SpaceX Is Buying Cursor"

vorticalbox — Tue, 16 Jun 2026 13:11:48 +0000

true, can't hurt though.

New comment by vorticalbox in "SpaceX to buy Cursor for $60B"

vorticalbox — Tue, 16 Jun 2026 12:58:03 +0000

Anthropic are already paying $15 b to space X for compute.

New comment by vorticalbox in "SpaceX to buy Cursor for $60B"

vorticalbox — Tue, 16 Jun 2026 12:56:18 +0000

its not just the models, their auto complete is actually really good. when you make a change it will give you "tab to next" which makes refactors super easy.

composer 2.5 is also a very decent model, it go 90% of my AI tasks using it now.

New comment by vorticalbox in "I indexed 669 GB of my GoPro videos using my M1 Max computer and local ML models"

vorticalbox — Sun, 14 Jun 2026 18:21:31 +0000

Vision still works perfectly fine in abliterated models.

New comment by vorticalbox in "Open source AI must win"

vorticalbox — Sat, 13 Jun 2026 12:00:32 +0000

where they came from doesn't matter so much so long as they are open weights and I can run on servers I control.

New comment by vorticalbox in "Show HN: Script to bulk delete Claude chats from the web UI"

vorticalbox — Fri, 12 Jun 2026 19:25:33 +0000

the request still went to a server you have no control over.

For all we know these temporary chats are saved and linked to your account but just don’t show up in the UI.