vidarh

11 ms·

241.
by vidarh 3mo ago
I think not looking at the original X source code is probably a good thing. The protocol is reasonably well specified, and you can further use the XML specification used for XCB. Most of the requests are pretty simple. For some of the trick
242.
by vidarh 3mo ago
Yeah, I've had it work on an X11 server using a Ruby X11 protocol implementation instead of libX11, and it just rushed ahead and added support for a bunch of missing requests and responses. None of that is hard - it's all very wel
243.
by vidarh 3mo ago
It's relatively recent, I feel. Claude used to really struggle with writing asm not that long ago. But the last 6 months or so it's done great. It's also far better than me (as someone who has done assembler since the Commo
244.
by vidarh 3mo ago
There are 3-4 typical different approaches to rendering text on X. If it's not yet (or not correctly) implementing one or two of them, that'd explain it. EDIT: Testing with xtruss shows st uses RenderCompositeGlyphs8(), part of th
245.
by vidarh 3mo ago
It's market cap is still closer to 2 trillion than 1.
246.
by vidarh 3mo ago
Yeah, I have a mostly-functioning X11 server in Ruby, myself (not anywhere public yet). Isene and I have relatively similar philosophies on this, except I have Claude burning tokens on optimizing and fixing my Ruby compiler now because I st
247.
by vidarh 3mo ago
The point of the hardware comparison was not to argue that the hardware should be optimised, but to point out that it's ludicrous to point to the scale of the difference as evidence it's not an engineering gap when we have a his
248.
by vidarh 3mo ago
Typical storage densities have improved by several magnitudes per decade for decades. A mere 3 magnitudes of gap doesn't just seem like an engineering gap, but a relatively small one.
249.
by vidarh 3mo ago
It was the best-selling single model of home computer for a very long time (relatively speaking in a very fast-moving field) in an era where most new models ditched compatibility, and in some countries it totally dominated. Where I grew up
250.
by vidarh 3mo ago
Granted I've only used it for a few hours, but to me K3 still appears below Sonnet, even below 4.5. I have their highest subscription, so it's not that I can't find uses for it and it has sides to it I really like, and it per
251.
by vidarh 3mo ago
I'll say after having it run for a few hours that I still don't feel it matches even Sonnet. It still does a lot of back and forth that feels dumb, but it's possible this is in effect Anthropic tricking us by hiding the full
252.
by vidarh 3mo ago
And yet a huge proportion of people are using LLMs daily. That people don't want data centers in their back yard does not translate to opposition to LLMs. Anyone who hopes it is will be setting themselves up for disappointment when peo
253.
by vidarh 3mo ago
My point is that you don't need a special purpose one to achieve this.
254.
by vidarh 3mo ago
You don't really need a style model - current models are very good at doing "style transfer" of a model text onto whatever it has written if you just have it do it chunk by chunk. It takes more to prevent it from being detect
255.
by vidarh 3mo ago
Pangram won't know how much AI written text they fail to detect, though, and detectors is a great tool to adjust methods of generating less AI-sounding text.
256.
by vidarh 3mo ago
> and we've seen the public around the world overwhelming reject LLMs Have we? Where do you see people rejecting them?
257.
by vidarh 3mo ago
KDP now has a limit of 10 books per week to limit it unless you can convince them you're a legitimate publisher, and e.g. Draft2Digital does not submit ebooks to Amazon any more for the same reason. But you can still offer paperbacks v
258.
by vidarh 3mo ago
It's a very common thing. A very unscientific quick search in Google Books led me to the oldest example I could find in a minute or two being a mention in The Dublin Review in 1844 reviewing a book on the correspondence of Edmund Burke
259.
by vidarh 3mo ago
> I thought this discussion is more about somehow aligning pipes over multiple lines in existing console emulators(impossible) I am aligning pipes over multiple lines in my existing console emulator. It's not just not impossible, bu
260.
by vidarh 3mo ago
Moonshot has subscriptions maxing out at $199/month. Not home so not had a chance to see if K3 is included yet. EDIT: Just switched my Kimi-CLI session to K3 and resumed my ongoing /goal... Will be interesting to see if I notice a
261.
by vidarh 3mo ago
No, you don't need to pick a fixed width font for that. You will get the best results with one, but rescaling the glyphs works just fine. I've written a font renderer. It's not hard. In fact my terminal, using said font rende
262.
by vidarh 3mo ago
You can still handle this well enough in the renderer, so I'd still consider this a renderer bug - e.g. my terminal emulator will scale any glyph that exceeds the bounding box as defined by the expected number of cells for a given char
263.
by vidarh 3mo ago
No, she's pointing out that when you have a dierse prize committee, you end up with more diverse results. > For example, apparently no white men born in the last 40 years have published literary fiction in the New Yorker. [1] How ma
264.
by vidarh 3mo ago
I think this is the case more often than people want to admit. A lot of genres demand the expected and often demand even the cliche and you better have damn good reasons for straying from it. Usually it takes a lot more skill than most
265.
by vidarh 3mo ago
That's pure generic "I'm getting paid by the word" article writing. If it reads to AI to you it's in part because AI has been trained on a billion articles like that.
266.
by vidarh 3mo ago
I overheard a cab driver being scammed by someone claiming to be HMRC (UK version of the IRS), and when he asked to call them back they managed to convince him to call them back on Viber , and he was about to when I intervened and pointed
267.
by vidarh 3mo ago
I rather think the surprise is a result of technical users wildly overestimating how obvious these markers are to people.
268.
by vidarh 3mo ago
I worked on a fine-tuning projects where the ultimate customer was a major provider, and their style guide at the time required us to fail any response that used "I". I've noticed the same as you, so I guess that is out the w
269.
by vidarh 3mo ago
Consider each LLM as one personality. We're getting corporate bullshit from a large number of different personalities, so even when they are similar they are still filtered through many different lenses. Most of us are only dealing wit
270.
by vidarh 3mo ago
Re: style-transfer, I simply point it at examples of my writing and asks it to use sub-agents and review paragraph by paragraph for consistency with my writing-style. I do the "paragraph by paragraph" thing because the main agent

More ›