I have found that you often have to provide some hints to the LLM. Also, providing server logs from production can also greatly assist it in finding performance issues. I have had pretty good success with this, and it has saved me a lot of time. Just be extremely scrutinizing about the solutions that it proposes, because those are frequently suboptimal.
What's the point in discourse if not to change the other person's mind? Triggering the limbic system of the person you are talking to is the fastest way to ensure they won't be able to engage with their PFC and actually hear and consider what you're saying. If the point is just to feel better about how righteous and right you are, then by all means proceed with your approach. But if the point is to influence somebody's views, then you are self-defeating in your approach.
Personally, I think federal officers have executed law abiding citizens. But if I start out by screaming "The Nazis have control of our government and are executing innocent people in the streets!" then not only have I closed my own mind to potential challenges to my views (which is at best hypocritical to expect the other person to be open-minded when I am not myself open-minded), then we get nowhere and just come away hating each other and thinking the other person is crazy. Worse, it poisons the well so the future reasonable person is immediately written off with guilt-by-association (person A was crazy and person B shares a view with them, therefore they must be crazy too).
> What's the point in discourse if not to change the other person's mind?
That was a question made at one of those public debates that the Oxford University likes to organise, and I think the answer is right on point: the purpose of discourse is to let the audience (or readers) reflect on an opinion, which takes time. It's *almost never* to change the opinion of the person you're debating. It's a given that most people that do like to engage in debate or public discourse are the kind of people that are unlikely to change their minds, and if ever they do, it won't be on the spot.
Ah, yeah that's fair since we're talking about moderating online discussions which are accessible for the public. Although I think the principle still stands for people who aren't approaching the discussion from a principle of neutrality. The people in the audience that you want to change the minds of will react similarly to the way I described, so you might get a small percentage of open minded people but you limit your reach. The extremity of the position also tends to resonate poorly with moderates/undecideds, so I would still suspect that a more reasoned, logical argument would be more effective with the audience. That said though, you make an excellent point.
I understand your point which sounds reasonable for a lot of debate, but the counter argument would be that in some situations you are normalizing both sides, when one side is not acting in good faith and is on the wrong side of history. Examples being Southern slave holders, Putin's invasion of Ukraine, fossil fuel interests regarding climate change.
If one did live under Nazis German rule, would it have been wrong to scream, "The Nazis have control of our government and are executing innocent people in the streets!"? At that point you're trying to wake the public up to do something about it, not sit down and have a debate over Goebbels latest speech with some fence sitter who can't decide whether Hitler has gone too far.
Indeed. A few years ago I spent a lot of time "fact checking" things, and it's nearly impossible because there is way more speculation/interpretation of "facts" than most people think. Misleading headline writing makes this even worse because a lot of people don't read beyond the headline, or if they do they interpret the factual body of the article through a lens framed by the headline. The NY Times are exceptionally good at this. Read the article and it's factually correct, but different interpretations and the subtle insertion of opinions (often through headlines) . I'm not trying to shit on NYT here. NYT is still among the best sources, despite their imperfections. But it illustrates well the challenge.
As much as I love sarcasm that is done well, I do find that it translates very poorly to written text unless explicitly noted with /s or something like that. Even when annotated, it's extremely rare that a sarcastic comment actually furthers discussion or makes a meaningful point. If a person is using sarcasm, odds are pretty high that they aren't engaging substantively anyway. Given the difficulties with detection (which even many humans fail at) it seems like trying to detect sarcasm would just make the tool a lot less useful and would be mostly antithetical to the project goals anyway.
That would shut out most people working for big corp, which is probably a huge percentage of the user base. It's dumb, but that's just the way corp IT is (no torrenting allowed).
It's a sensible option, even when not everyone can really use it. Linux distros are routinely transfered via torrent, so why not other massive, open-licensed data?
Oh as an option, yeah I agree it makes a ton of sense. I just would expect a very, very small percentage of people to use the torrent over the direct download. With Linux distros, the vast majority of downloads still come from standard web servers. When I download distro images I opt for torrents, but very few people do the same
> very small percentage of people to use the torrent over the direct download
BitTorrent protocol is IMO better for downloading large files. When I want to download something which exceeds couple GB, and I see two links direct download and BitTorrent, I always click on the torrent.
On paper, HTTP supports range requests to resume partial downloads. IME, it seems modern web browsers neglected to implement it properly. They won’t resume after browser is reopened, or the computer is restarted. Command-line HTTP clients like wget are more reliable, however many web servers these days require some session cookies or one-time query string tokens, and it’s hard to pass that stuff from browser to command-line.
I live in Montenegro, CDN connectivity is not great here. Only a few of them like steam and GOG saturate my 300 megabit/sec download link. Others are much slower, e.g. windows updates download at about 100 megabit/sec. BitTorrent protocol almost always delivers the 300 megabit/sec bandwidth.
With Linux distros they typically put the web link right on the main page and have a torrent available if you go look for it, because they want you to try their distro more than they want to save some bandwidth.
Suppose HF did the opposite because the bandwidth saved is more and they're not as concerned you might download a different model from someone else.
Testing. If you share something you've tested and know works, that's way better than sharing a prompt which will generate untested code which then has to be tested. On top of that it seems wasteful to burn inference compute (and $) repeating the same thing when the previous output would be superior anyway.
That said, I do think it would be awesome if including prompts/history in the repos somehow became a thing. Not only would it help people learn and improve, but it would allow tweaking.
It's gotten more popular for sure, but it's definitely been around a long time. Even just on HN there have been conversation about gdb tui ever since I've been here (starting browsing HN around 2011). For anyone who works in embedded systems it's a very common term and has been since I got into it in 2008-ish. I would guess it was much more of a linux/unix user thing until recently though, so people on windows and mac probably rarely if ever intersected with the term, so that's definitely a change. Just my observations.
I'm not GP, but I have backups, plus I always make sure I've committed and pushed all code I care about to the remote. I do this even when running a prompt in an agent. That goes for running most things actually, not just CC. If claude code runs a git push -f then that could really hurt, but I have enough confidence from working with the agents that they aren't going to do that that it's worth it to me to take the risk in exchange for the convenience of using the agent.
You can roll a script to do this, something that would consume a mic from Pipewire when triggered and then push results to clipboard. With a Parakeet ONNX model in between.
I had cause to do the the opposite: Hotkey -> clipboard TTS
My ~1.7% WER and faster than realtime processing in my application make it more than adequate. My application is multi-speaker with WPM rates >300 for long durations.
reply