vorticalbox

7 ms·

91.
by vorticalbox 4mo ago
where they came from doesn't matter so much so long as they are open weights and I can run on servers I control.
92.
by vorticalbox 4mo ago
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.
93.
by vorticalbox 4mo ago
I switched to oMLX today from lm studio. Really nice but I have found qwen3.6 sometimes failing to call tools correctly.
94.
by vorticalbox 4mo ago
when i was a teen the "sleep" button was just right of the enter key so if you slipped, which i often did, night night computer.
95.
by vorticalbox 4mo ago
I run a few local models for different things. I find Gemma 4 great for writing but qwen better for coding. I tried the same prompt on gemma4 and qwen 3.5 and Gemma consistently failed to call the multi line edit tool.
96.
by vorticalbox 4mo ago
for the last few weeks I have been using composer 2.5 (cursors fine tune of kimi 2.5) and honestly i don't see it worth the price to use 5.5, opus or sonnet any more. for almost all the tasks i have given it, it has handled it perfectl
97.
by vorticalbox 4mo ago
if an account with the ability to push to the repo was taken over, there wouldn't be any PR review.
98.
by vorticalbox 4mo ago
New LLM benchmark?
99.
by vorticalbox 4mo ago
Not hard, one could build an application that listens on common software ports and simply returns 200 for every request it gets. Not sure how it would benefit you telling some website you run all the software.
100.
by vorticalbox 4mo ago
annoying true, but no reason they couldn't group read/write into the same prompt.
101.
by vorticalbox 4mo ago
with ACP[0] they can control one another. [0] https://agentcommunicationprotocol.dev/introduction/welcome
102.
by vorticalbox 4mo ago
i think the issue is that you can still have these all under the accessibility api but why not break that down more accessibility.screenshot accessibility.paste and whatever else there is. that completely removes the issues for apps like th
103.
by vorticalbox 4mo ago
Their job is to generate text if that text is good or bad they are functioning as intended.
104.
by vorticalbox 4mo ago
they have a webGPU demo [0] at 4 steps it takes 7 seconds to generate an image on my M4 https://huggingface.co/spaces/webml-community/bonsai-image-w...
105.
by vorticalbox 4mo ago
some of the bench marks i have seen on also include cost where one scan of the codebase cost tens of thousands of dollars. this one [0] notes one run cost $20k to run but another cost $50. [0] https://red.anthropic.com/2026&
106.
by vorticalbox 5mo ago
async/await didn't hit javascript until 2017. in 2015 you picked between function foo(cb) { cb(42, null) } or function foo() { return new Promise((resolve, reject) =>{ resolve(42) }) } the second allowed for .then and .c
107.
by vorticalbox 5mo ago
Same but I have seen it try and change my tests because it decided that it’s code was correct and my tests were incorrect. I saw it’s thinking tokens said something along the lines of “I have implemented it correctly but the test is failing
108.
by vorticalbox 5mo ago
I don't fully get the argument about errors. in rust say a function returns Result<T, E> so either the we get a result all an error how is that different from (int, err) in go? do you not still need to handle the error? in go you
109.
by vorticalbox 5mo ago
I don’t consider them a problem at all. It’s how browsers have worked for a long time. My point was at the time when deno came out it was completely different to node with its imports and that meant a lot of existing packages just didn’t wo
110.
by vorticalbox 5mo ago
I think the main issue was when deno first came out it used urls for imports then later added support for npm. By then bun was already a thing and just ate into its share.
111.
by vorticalbox 5mo ago
Yes https://cursor.com/cli
112.
by vorticalbox 5mo ago
I’ve seen Claude write a one line Perl script to pull items from a json file.
113.
by vorticalbox 5mo ago
Not in the UK sadly, ADP is disabled.
114.
by vorticalbox 5mo ago
I do this too I never made the connection.
115.
by vorticalbox 5mo ago
If you build a small ui that lets you edit the models response too it’s pretty funny to do. It sees that it “said” it and gets very confused.
116.
by vorticalbox 5mo ago
In the video he said that the courts ask "who is the warrant for" and he replied "no one", but surely one could also look up the number plate and find it that way?
117.
by vorticalbox 5mo ago
codex will actually help you look but it will refuse to actually try and exploit it. it won't for example create a POC python script that you normally would use to prove the issue.
118.
by vorticalbox 5mo ago
I am software dev and i was doing a security check on my own application (work) I was running in localhost and gave it access to the code. every single model refused to attempt to run any sort of test to check if it was a n issue other than
119.
by vorticalbox 5mo ago
It is learning though. It’s not just copying the code. Code gets turned into tokens and then it learns the next most likely token. The issue that I see most people talk about it the scale at which is learnt. A human will learn from other pe
120.
by vorticalbox 6mo ago
This reminds me of https://dnhkng.github.io/posts/rys/ David looks into the LLM finds the thinking layers and cut duplicates then and put them back to back. This increases the LLM scores with basically no over hea

More ›