gaigalas

5 ms·

1.
by gaigalas 4d ago
Open source is Senna on a Toleman in Monaco '84 and Anthropic is Prost cancelling the race he was going to lose blaming safety concerns.
2.
by gaigalas 5d ago
Yeah, "Your anti-virus is outdated, click here to update" popups mimicking the OS were a real issue. That's also why many browsers also don't allow popups without an URL bar now. For some time, Internet Explorer even had
3.
by gaigalas 5d ago
My point is that opinionated software is a magnet for supremacist types and ideologies. It's a nest for their "vision" or whatever. We're already way past the point of pointing out his character flaws. I want to also poi
4.
by gaigalas 5d ago
> There’s also nothing wrong in more opinionated software Except the fact that it's a breeding chamber for those "unity and identity" types of people/ideologies/projects? Embracing diversity is messy , and it&#x
5.
by gaigalas 6d ago
They exist but don't actually grab system colors (to prevent pages spoofing the system look). --- Shameless plug: In neutrino (my webview wrapper), we grab those and inject them as prefixed CSS variables (--neutrino-NameOfSystemColor).
6.
by gaigalas 6d ago
Missing the whole IE quirks mode vs standards mode thing, but nice. Some of those hacks were specifically for IE, and there was a debate on staying quirks vs going standards mode. Each had its own drawbacks and choices.
7.
by gaigalas 6d ago
Thank you.
8.
by gaigalas 6d ago
sources?
9.
Coletivo – A place for AI agents to collaborate via MCP(coletivo.alganet.dev)
1 points by gaigalas 6d ago|0 comments
10.
by gaigalas 13d ago
Yes, yes. Let's do that with testing and QA too. In massive scales. Humans are too slow. I want to see what happens.
11.
by gaigalas 13d ago
I think we should let reviews go. The next in the list are tests and CI, they also slow down delivery, and should go too. Then, we'll have the full speed and the consequences of that. I want to see it happening on a massive scale.
12.
by gaigalas 18d ago
Yep, but that's not changing the quality of the model. It's not an optimization in any sense (and it's a hit on productive workflows, possibly). This is also likely to stop working as censoring moves to the training data sour
13.
by gaigalas 18d ago
I think those are mostly vapor that runs on the small culture of "models should not be censored" thing. But from my experience, they unlock nothing meaningful. Fine-tuning is great for really small models on specific applications,
14.
by gaigalas 18d ago
Optimization on a idiosyncrasy. The same thing that makes Claude repeat "That was the most important thing you said in this whole conversation" is what makes grug speak optimize on token usage. Real humans get non-primary informat
15.
by gaigalas 18d ago
It's not exactly a joke, it does reduce the amount of tokens. However, it does not improve performance (fine tunes are finnecky things, hard to get one right).
16.
by gaigalas 19d ago
2025?
17.
by gaigalas 22d ago
Depending on the cause of the current issues, that move would likely cause more harm to paid services than good. Their last postmortem made clear that their challenges are operational. Scale puts pressure on operation, but it's not wha
18.
by gaigalas 24d ago
It's down. It seems that Fable is recovering, which is interesting (likely a priority, or more stable, or who knows).
19.
by gaigalas 29d ago
If we're switching, we should probably go to fossil instead.
20.
by gaigalas 29d ago
I lived in a short period of time in which QA was really good. Early Jenkins era, before GitHub. People engineered a lot of ingenious stuff to prevent bugs. One team I worked with had tests for the product we made ranging from IE6 to IE11,
21.
by gaigalas 29d ago
This age: we made the thing that codes faster before we made the thing that does QA faster.
22.
by gaigalas 1mo ago
Weights to say "no" reliably might be another order of magnitude (or two) compared to what LLMs have today.
23.
by gaigalas 1mo ago
Maybe I'm not seeing the speed improvements because it's a dense model, and I'm comparing it to the 3.6 MoE with MTP. Likely, I'm attributing the speed I see to something else. Honestly, if MTP is available, I don't
24.
by gaigalas 1mo ago
I don't see an MTP entry on Unsloth though. Maybe it's not available in a lower quant I need for my poor GPU.
25.
by gaigalas 1mo ago
I have an idea for AI companies for tremendous scale, with privacy and all nice things accounted for: - Unstarve the GPU and RAM consumer market. - Let enthusiasts and volunteers quickly ramp up local AI. - Reap the results that the communi
26.
by gaigalas 1mo ago
Waiting for the MTP version to pop up on Unsloth. Speculative decoding makes a huge difference. Been running quantized 3.6 at 110t/s on a cheap 5060Ti and quite happy with it. If 3.8 improves on it, it would be awesome.
27.
Wasi-sh, run limited shell scripts on the browser(github.com)
3 points by gaigalas 1mo ago|1 comments
28.
by gaigalas 1mo ago
This is the engine behind my blog's terminal view, which I just open sourced. More info: https://alganet.github.io/blog/2026-07-21-00-Bash-for-Browse... It is fairly fast, runs on node and browser, and can serve a
29.
by gaigalas 1mo ago
Go hard boring on 80% of what you do. Go hard exciting on the 20% else. Don't let one affect the other.
30.
by gaigalas 1mo ago
If someone is ignoring basic instructions like "don't worry about the scriptures", then it doesn't matter anyway. Basic text interpretation skills are missing in the interaction.

More ›