solarkraft

7 ms·

61.
by solarkraft 2mo ago
I agree. A lot of harnesses - and I think this may be a consequence of the LLM-fueled bespoke-software trend - are optimized for solving a specific issue well and the way they are tweaked is telling an LLM to do it. This resolves the need f
62.
by solarkraft 2mo ago
Baking them into the model and having them apply this strongly is hard and resource intensive, as far as I am aware. Having them in context is super easy and cheap. It is trivial to change and is 100% cacheable.
63.
by solarkraft 2mo ago
Right, so it's as correct as me claiming it to be an E-Mail sorting model. It may be good at that, but that's not its primary purpose.
64.
by solarkraft 2mo ago
If there is a repeated, ongoing problem with a lot of people using the devices to crush puppies, the manufacturers would be expected to come up with ways to make this less likely to happen (safety measures etc). And if that didn’t help, the
65.
by solarkraft 2mo ago
Waymos don’t precisely harass women and upload the footage to their instagram feed
66.
by solarkraft 2mo ago
I feel like we’ve had this discussion before. From what I remember, specialized models rarely do that much better than general ones, hence no mode Codex models.
67.
by solarkraft 2mo ago
Wow, Meta is back (at least for now)! I like this class of model. Multi-token prediction makes it viable to run dense models at not-too-far-off speeds as MoE models with much better intelligence. The submission’s title (open weights 30B l
68.
by solarkraft 2mo ago
Oh, that sounds related to what I experienced. I recently had another look at Zed and tried to use it to full-text search through my workspace consisting of symlinked folders. Relevant things would just now show up, cmd+P also wouldn’t find
69.
by solarkraft 2mo ago
Oh, absolutely. That is the hot path.
70.
by solarkraft 2mo ago
> That said, it is noticeably slower than Zigbee, but we’re talking 50ms > 250ms. Which is more than acceptable. I’m taking your “but” to mean “it is acceptable. I would at least deem it noticeable to wait a quarter second for my lig
71.
by solarkraft 2mo ago
Yes.
72.
by solarkraft 2mo ago
Yes
73.
by solarkraft 2mo ago
No, it means that I’m building a fast alternative to the stock solid based web UI. Pi (isn’t the “mono” just about it being a monorepo?) has no features and no defined server-client API.
74.
by solarkraft 2mo ago
The open models are good! Even when hosted, they buy you a lot of portability.
75.
by solarkraft 2mo ago
Thank you for your work on Matrix. I believe that the world needs an interoperable messaging standard just like we have E-Mail, which is why I am so interested in Matrix becoming something I can tell people to use with a good conscience. I
76.
by solarkraft 2mo ago
Yes, this would be acceptable too, if that is all we had. We would build systems do deal with this unreliability, just like we have with humans.
77.
by solarkraft 2mo ago
The website is also so hilariously enshittified that the “enshittification” text barely peeked through the ads and other popups.
78.
by solarkraft 2mo ago
Seeing such issues with Matrix and other open, common-good projects, I get the feeling like they just aren’t architected for good everyday-user UX. Matrix is making real efforts to change this with Matrix 2.0, but still seems to suffer from
79.
by solarkraft 2mo ago
Same for multi-account support in Element Desktop. However: The new breed of clients (Element X) has this, along with many other improvements. This is probably what the happy user featured in the article was using. They recently (in the las
80.
by solarkraft 2mo ago
> and desktop access is actually routed through it I think this is not true anymore. I’ve had it working with while my phone was off.
81.
by solarkraft 2mo ago
I can’t think of any example in which it has gone like this, but some in which it has: The most prominent in my mind would be software transitioning from “of course you get the code so you can extend it yourself” to what we have today (plea
82.
by solarkraft 2mo ago
This is an important article. I hadn’t realized it was already getting this bad. Like a frog enjoying a nice warm bath ... > Most people do not switch their operating system or phone provider every week either. But even if you do not uti
83.
by solarkraft 2mo ago
> One observation: ask an LLM to write a skill, it tends to make them overly verbose and prescriptive. Often, something briefer and human-written actually works better Incredibly so. All too often a “hey, you shouldn’t have done that her
84.
by solarkraft 2mo ago
They have no (other) equivalent to nano, so it makes sense that it’s much cheaper now. It may have been better, but it was also hell of a lot more expensive.
85.
by solarkraft 2mo ago
What an incredible project. I also find it tragic how much Apple hates people running software on their devices that makes them vastly more useful.
86.
by solarkraft 2mo ago
Cranking up the reasoning effort seems be a trend recently and to me it looks desperate, like cranking up the wattage of a processor when you can’t get other, better gains. Like yeah, you made it perform better, but I could’ve set any other
87.
by solarkraft 2mo ago
I get your reading. My interpretation is that they’re claiming enshittification and that the local models don’t have that. It’s an understandable view, but I’d be astonished by any local model processing very long context better than any fr
88.
by solarkraft 2mo ago
Oh, great point!
89.
by solarkraft 2mo ago
Also see the german parliament’s laws repository on Github: https://github.com/bundestag/gesetze
90.
by solarkraft 2mo ago
The opposite applies to my use. Humans are generally great at predicting where moving things will land - we evolved for that. It feels great flicking the cursor around and stopping it at the right point. It’s just like scrolling a list on y

More ›