rurban

15 ms·

481.
by rurban 8mo ago
It not hard since opus or gpt-5.4 anymore. I check most of my Makefiles with them to update wrong deps.
482.
by rurban 8mo ago
Wait for the great new times when an AI will certify aerospace, automotive and medical SW. Waiting for that. It will be 1000x better and faster than the existing processes
483.
by rurban 8mo ago
I have to maintain our GPU's. Generally the worst parts are the watercooling pressure, the HVAC, and the power. I can run it stable only at 300W per CPU, the normal max is 310W. Now with throttling to 300 it's stable for a year, b
484.
by rurban 8mo ago
The zero regression rate graph at the end is exactly my experience. Only Opus is useful right now, the rest are juniors.
485.
by rurban 8mo ago
Tuning guitar strings by hand is trivial. Try to tune a piano. There have up to 3 strings per key, and quite a lot of keys. Cannot be perfect, but should sound "warm" enough.
486.
by rurban 8mo ago
We did run it locally on a free H100, and it performed awfully. With vLLM and opencode. Now we are running gpt-oss-120b which is better, but still far behind opus 4.6, the only coding model which is better than our most experienced senior d
487.
by rurban 8mo ago
I talked to my tax office, and was even selected for a check. No problem at all.
488.
by rurban 8mo ago
I needed one step only, and certainly no tax advisor. Most costly is the health insurance, but you can skip the Gewerbe Anmeldung and taxes and insurances with doing systems SW. You only need a GmbH if you invest money, don't do that.
489.
by rurban 8mo ago
I hate HW soo much. To revise the biggest problems in computing, beside out of tokens: HW bugs
490.
by rurban 8mo ago
The question is still: Does it make your code better or worse? Only Opus makes it better, the rest worse. That's the treshold
491.
by rurban 8mo ago
Biggest problem right now in computing: Out of tokens until end of month
492.
by rurban 8mo ago
I've tried Qwen3.5-35B-A3B locally on our H100's, and it was horrible. Much more halluzinations than anything else. The big qwen coder would take 4 GPU's which is a bit too much. Trying GPT-oss-120b tomorrow. For now we are h
493.
by rurban 8mo ago
No, the (github issue) ID is enough. For my AI coding sessions I just point opencode to the issue. It does a plan, (build, ie) implements and tests the plan, and commits it. For reference you always have the issue, revise the issue when som
494.
by rurban 8mo ago
In Europe this guy stands for vulgar and racist neofascism
495.
by rurban 8mo ago
That's interesting. We do fine-tune ours, but we do vision.
496.
by rurban 8mo ago
I switched from duckduckgo to seek ninja about a year ago, and it's much better. Kagi was much better, but I certainly won't pay for search nor AI
497.
by rurban 8mo ago
Why not pcre which has more features, better performance and is battle tested, compared to that minimal new matcher without any proper locale info? They must really love those experiments.
498.
by rurban 8mo ago
I don't know of any good backbone service or library with 5K+ stars, only npm or python hyped bozos. There is certainly not any competence in github stars. Aren't any GNU libs maintained on GitHub? No, they are mostly not. Just ve
499.
by rurban 8mo ago
The headlines in Europe are that Israel is carrying out preventive strikes, the USA is helping. And that's certainly the deathbed of any hopes to a mullah regime change. They will come out stronger than before.
500.
by rurban 8mo ago
I gave lots of prolog rules to analyze log files of a complicated distributed system with 20 realtime components to find problems and root causes. Worked really well. In 2008 or so Cannot believe that LLM are that useful. When ever a compo
501.
by rurban 8mo ago
Typo: (CoreOS and Fedora Silverblue) are the bleeding edge of immutable distros. Those mentioned are just users.
502.
by rurban 8mo ago
Wealthiest nation? More like unwealthiest of all nations. U. S. by far the largest current-account deficit (over $1.2 trillion). U.S. has the largest goods trade deficit (over $1 trillion).
503.
Open-Source LLMs in 2026: A Complete Guide for AI Developers(vertu.com)
1 points by rurban 8mo ago|0 comments
504.
by rurban 8mo ago
I didn't get the top4grep tool to download most paper indices. This looked useful, but I only got results for a few conferences. But the link to https://decompilation.wiki/#decompilation-paper-list got me to a much bet
505.
by rurban 8mo ago
Nonsense. You just need to make building the gtk2 unit optional, so that the distros can still build it. Almost no one needs gtk2, just Lazarus. Usually debian maintainers are happy to patch the build system to do that. They got a bad one.
506.
by rurban 8mo ago
Nonsense. He dislikes better security reports because they are verbose? I got many LLM generated security reports lately, and they were just better. Way better. Detailing every step. The better the report, the less time the maintainer needs
507.
by rurban 8mo ago
That's a user pipeline problem. If you just check confusables without any tr39 algo you will be disappointed also. I had to patch confusables for the C/C++ proposal for about 10 characters also. "...the default confusables li
508.
by rurban 8mo ago
SBCL doesnt know when it's running low on available heap space? clisp uses libsigsegv, so it knows when to garbage collect really, and when it's not so needed.
509.
by rurban 8mo ago
Also racket since then was ported to use the fastest known scheme, chez, from the quite simple and slow mzscheme.
510.
by rurban 8mo ago
Missing: Have the minor padding changes by NIST (=NSA) to Keccak for SHA-3 really not impacted security? https://crypto.stackexchange.com/questions/15727/what-are-th...

More ›