dsrtslnd23

4 ms·

1.
by dsrtslnd23 20d ago
252GB of HBM3e at 100k vs. a multi A6000 96GB setup. GB300 seem expensive in comparison at ~$100k. Am I missing something?
2.
by dsrtslnd23 21d ago
are we sure humans have that choice?
3.
by dsrtslnd23 23d ago
I also built a custom harness tailored to my needs using pi.dev underneath. From time to time I use the 'official' harnesses of openai/anthropic but can't confirm that they are much better adapted to that harness.
4.
by dsrtslnd23 1mo ago
exactly - this should be trivial.
5.
by dsrtslnd23 1mo ago
yes would also be interested in that - using knowledge distillation and other special knowledge sources to post-train on top of an open model like GLM-5.3. I was wondering that when Elon Musk tweeted "Specialist AI’s (single language,
6.
by dsrtslnd23 1mo ago
I think it really depends - for a lot of things outside of coding and general knowledge tasks even the best models (fable 5 etc.) are not good enough yet: e.g. CAD, PCB design (though getting there on PCB design), ...
7.
Nvidia AVO scores 100% on the ARC-AGI-3 interactive reasoning benchmark(twitter.com)
71 points by dsrtslnd23 1mo ago|41 comments
8.
by dsrtslnd23 1mo ago
will this be open weights?
9.
by dsrtslnd23 2mo ago
I hear that often but to me it does not feel like it. I built my own framework around pi.dev harness and run all kind of different LLMs with it. Sometimes also use the vendor harnesses and they don't feel better adapted.
10.
by dsrtslnd23 2mo ago
maybe a quantized version on a GB300 would work? unsloth hopefully working on it.
11.
by dsrtslnd23 3mo ago
Seems that this will not be open weights?
12.
by dsrtslnd23 3mo ago
With all the technological advances, why are wind tunnels still widely in use today (instead of pure computational approaches)?
13.
by dsrtslnd23 3mo ago
what is the current statement on Oauth Claude use with Pi. I remember there was some back and forth from Anthropic? I currently use my Claude sub with Pi.
14.
by dsrtslnd23 3mo ago
I agree - OpenMover has that approach. Use a COTS mover and change the brains only https://openmower.de/
15.
by dsrtslnd23 3mo ago
I agree, but there are use cases for the 'best model' other than converting your 1975 stuff to rust: for use cases where LLMs are just getting started to be useful I really want to use the current 'best' model: e.g. CAD,
16.
by dsrtslnd23 3mo ago
i thought they did allow third party tools - or did they revert on that again?
17.
by dsrtslnd23 4mo ago
looks like I need a GB300 workstation
18.
by dsrtslnd23 4mo ago
I feel like polling mail 30+ years ago on ISDN + zipped mail file from my fido net node was faster than IMAP on my 1 gbit connection now.
19.
by dsrtslnd23 4mo ago
i find it very confusing - there were reports of accounts being banned due to it - then allowed again - what is the current state? I would like to use it with pi.dev harness.
20.
by dsrtslnd23 4mo ago
Didn't they allow using oauth in custom harnesses for personal use (e.g. pi.dev)?
21.
by dsrtslnd23 6mo ago
actually gpt-image-2 is VERY flexible with the resolution. You can use arbitrary resolution within the max pixel budget.
22.
Ask HN: Why isn't FPGA adoption exploding now that LLMs can write HDL?
2 points by dsrtslnd23 6mo ago|2 comments
23.
by dsrtslnd23 6mo ago
turn your volume down before opening...
24.
by dsrtslnd23 7mo ago
you mean this one? https://github.com/vercel-labs/agent-browser
25.
by dsrtslnd23 7mo ago
I think a terminal is still the perfect UI for coding agents on desktop systems. But now that I have running agents in background that only sometimes require some extra input on the go - I hate the terminal use on the go: terminus or other
26.
Ask HN: Any informed guesses on the actual size/architecture of GPT-5.4 etc.?
6 points by dsrtslnd23 7mo ago|0 comments
27.
by dsrtslnd23 8mo ago
Cadquery looks interesting. In particular STEP support compared to OpenSCAD. Thanks for mentioning.
28.
by dsrtslnd23 8mo ago
There's a whole spectrum forming here. Clacker News is the HN end, this is the 4chan end. I give it two weeks before someone launches a bot-only LinkedIn.
29.
by dsrtslnd23 8mo ago
currently using org-roam and was wondering about having something like this. Really cool!
30.
by dsrtslnd23 8mo ago
yes - indeed, would love to see generic IMAP support.

More ›