Rendered at 19:51:25 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
mentalgear 10 hours ago [-]
It's a nice idea, but I have mostly stopped using/installing any software that is not open-source a long time ago. So, please open-source it, especially if you want users to truly trust it.
Even then, I would recommend anyone to install (small to mid) browser extensions by cloning and inspecting the source and just then loading it yourself - if you don't know: any browser extension can read input/password fields across all site(s) you gave it access to (yeah, it's crazy but unfortunately true).
smusamashah 10 hours ago [-]
This could become/converted to a userscript making it easy to inspect and more cross compatible. It's very easy these days.
imiric 9 hours ago [-]
To be fair, the .xpi is just a .zip file, and the JS isn't minimized.
I vibe-converted it to a userscript[1], but it doesn't work for me in qutebrowser because it depends on profile data fetched from `https://hn-trustspark.com/alltrust.json`, which seems to be periodically updated by the author, and qutebrowser has limited userscript support. :(
It could probably be worked around by fetching the data externally, but I don't want to depend on hn-trustspark.com. It would be great if the profile updating tool could be published as well.
Anyway, hope it helps someone else :)
FWIW, after a quick review of the script, it looks safe to me. As long as you trust that the served profile data is correct, and don't mind leaking your IP to it.
Great idea and kudos to the author! We need more tools like this to help us deal with spam, and not just on HN, but everywhere.
Thanks! Nice, you found the alltrust.json file ha. Yes, a bg job running on an rpi leverages HN APIs and builds the alltrust file by the minute, for all "active" accounts. Technically fetching that data is all you'd need to make your own script/plugin.
It's centralized for a few reasons though, first being that client-side API requests would be discourteous to the APIs (flood/ddos), and a whole new level of error handling would be required. Shared IPs, like those in a tech company building, would easily and quickly reach the API limits. So that's the reasoning, if you're curious.
defrost 8 hours ago [-]
It's a good idea, provided nothing sketchy goes on either now or in a future update ... I'm playing about with it but will likely unload it soon enough.
One issue: New accounts with Zero submissions get a full green stack for "Submission Trust" (0 submissions in 0 days).
That comes across as three flat red lines and a singl tall green stack - not sure if that's the right message.
Submissions is an odd one ... those people that submit a story seen on the tech blogs once a day get bucketed together with habitual spammers of poor quality posts.
9 hours ago [-]
solaire_oa 4 hours ago [-]
Author here. This is a good callout, there are a few reasons why it's a plugin and not open source (yet).
First is that I didn't want to make a plugin in the first place, I wanted to make a bookmarklet, but HN's CSP policy was too strict. So that was a bummer.
Second is that I have very mixed feelings about open source these days, and so open-sourcing feels less and less like the sensible default state. One of the sibling comments here discovered the alltrust.json and vibecoded around it, which is really a case in point about why open sourcing feels like I'd be leaving myself "open" to be domineered (not just by users, but by bots and companies as well).
Third is that the system/plugin is partly LLM-assisted itself (even though the code is minuscule), and I'm self conscious of being a slop-slinger. Or at least, pushing up repos with LLM code just feels, idk... lazy and asymmetrical (despite this plugin having clear utility, which I think it does).
But it's completely fair to say "oh look, a plugin about trust that's closed source, how hypocritical." I get that. If there's enough interest I'll open source it, sure.
imiric 33 minutes ago [-]
Hey, that's a perfectly reasonable stance, and I can relate to it.
Apologies for working around it and putting the code out there against your wishes. If you check my post history, you can see how opposed I am to these new tools, and "vibe coding" specifically. In my defense, I really didn't want to spend a lot of time on this, and LLMs do a decent job at this type of mechanical conversion. And I really don't judge anyone for using them mindfully, as you've clearly done in this case. The code didn't read like slop to me, if it's any consolation. :)
Besides, this "closed source" criticism is really a non-issue in this case considering it's a browser extension with clear JS, which anyone can inspect if they were really interested.
Cheers!
EDIT: I've deleted the gist. :)
password4321 8 hours ago [-]
Wow "trust as in trust me bro", thanks for the heads-up. Only Y Combinator is allowed that level of access to HN karma metadata right now; what a gold mine.
bryanhogan 10 hours ago [-]
Interesting idea, wondering if the 4 bars are a good way to evaluate accounts.
Some more thoughts:
1. The website needs a description of what this project even is.
2. Open-source the code, especially since it's just for the community, I'd be vary of installing "some random" extension from the store.
3. The modal element in the iFrame on the website is cutoff on mobile widths using Firefox.
cpa 12 hours ago [-]
Interesting, but how is the trust signal measured? I couldn’t find this information
annie511266728 11 hours ago [-]
If it’s not transparent, it kind of defeats the whole idea of a “trust” signal — at that point it’s just a number you either believe or ignore.
xnx 8 hours ago [-]
"trust me bro" signal
cobertos 12 hours ago [-]
I'm pretty sure it's based off of the configuration in the green gear in the top right of the iframe. You can inspect the ways it's calculated from karma, comments, etc etc
CalRobert 11 hours ago [-]
If you look at the config it's based on karma, comments, submission rate, comment rate (optional), and account age (that is, if you trust it actually uses the config how it says)
angry_octet 9 hours ago [-]
There are some high karma accounts which make a great contribution, and others which ... don't. I'm doubtful whether karma or account age has a significant signal beyond anti-spam.
But it would be useful to know if I had up/down voted them significantly in the past.
krapp 9 hours ago [-]
The only thing karma reliably indicates is participation over time, the signal is too noisy for anything else. If anything high karma should be a red flag. The very best contributors here rarely comment because they have better things to do. It shows an 8/8 score for me and I doubt anyone would consider me a top tier high quality contributor.
A plugin like HN Comments Owl would be more useful IMHO.
mooreds 6 hours ago [-]
Site breaks for my account? Maybe I'm not trusted? But I see no sparklines next to my username when I visit https://hn-trustspark.com/ and have posts on the newest page.
mooreds 6 hours ago [-]
Actually, it did just blur out one of my submissions so maybe it does work and my other submissions are, as the kids say, mid.
solaire_oa 4 hours ago [-]
Oh wow, You're a human with a high submission rate! I assumed accounts like these were bots. I've seen a large number of accounts like yours in `/newest`, indeed, it's the reason I made the default demo penalize high submission rates. If you don't mind me asking, what are you submission habits? Do you just submit links you find interesting often? And does the karma rewards factor into your routine submissions?
(I'm not being facetious or accusatory, I'm genuinely interested learning how some of these high submission rates operate, since a lot look automated).
mooreds 2 hours ago [-]
haha, no worries.
> what are you submission habits? Do you just submit links you find interesting often?
Pretty much. I will often email myself links I think might be interesting, then batch up submissions. I also like to highlight friends' posts.
> And does the karma rewards factor into your routine submissions?
It's always fun to get karma, but I like the fact I can call attention to someone or something that deserves it more.
I also wrote and talked about my thoughts on HN and how to be a good community member a bit more:
I think friend/foe and trust signals should come from a user's voting. So I think it should operate transparently to the user and there should be a default shade-out option of bad actors, but with the option to view. So it's set-and-forget. On the backend - could you make it so if you've never visited HN before, and you install the plugin, the experience changes accordingly?
I had the idea to setup a forum somewhere else - this sort of functionality would come in handy as part of normal operation.
onli 10 hours ago [-]
Suggestion: instead of just popping in the computed graphic add a placeholder immediately - four low grey bars should work well, in the same width as the final sparkline graphics. That way the text on the page will not jump later when the final sparklines are ready.
croemer 9 hours ago [-]
Does anyone have an archive.is? Page hasn't been loading the 4 times I tried over the course of an hour.
Great idea! I've been running my own chrome plugin I made that is similar, but recently discovered this: https://oj-hn.com/ and intend to start to contribute to it as it's very good and hope it grows. Something like Sparklines would be great in oj also for us chrome users, I really like the Sparklines implementation.
Kkoala 9 hours ago [-]
This should have at least a basic description of how the trust signal works and is calculated
arjie 10 hours ago [-]
I considered something like this but I find the UX of information too noisy. In practice, I much prefer simply performing a ternary block/leave alone/highlight functionality. And it's comments that bother me more than links.
alex7o 12 hours ago [-]
Where can I see mine?
CalRobert 11 hours ago [-]
Install the plugin and then load up your profile page, or any page with ac omment from you (like this one)
You have an overall 7 out of 8.
no_shadowban_3 10 hours ago [-]
[dead]
vivid242 11 hours ago [-]
After attention, are we now in the trust economy… a brilliant idea for a plugin - thank you!
amelius 10 hours ago [-]
The problem is that I don't trust plugins.
Is there a way to do this from a bookmarklet?
solaire_oa 4 hours ago [-]
Author here. I tried a bookmarklet, that was my preference, but the security headers in HN were too strict. As far as I could tell anyway.
Lapra 8 hours ago [-]
The problem with having a public trust score is that it'll be gamed by bad actors A/B testing their bots surely.
layer8 8 hours ago [-]
I wouldn’t call four bars a sparkline.
Izmaki 11 hours ago [-]
Great way to gatekeep new contributors trying to be helpful with news. Run the link through URL Scan if you’re paranoid…
hluska 8 hours ago [-]
I wish I wouldn’t have read this comment from you, so you may be onto a feature.
Even then, I would recommend anyone to install (small to mid) browser extensions by cloning and inspecting the source and just then loading it yourself - if you don't know: any browser extension can read input/password fields across all site(s) you gave it access to (yeah, it's crazy but unfortunately true).
I vibe-converted it to a userscript[1], but it doesn't work for me in qutebrowser because it depends on profile data fetched from `https://hn-trustspark.com/alltrust.json`, which seems to be periodically updated by the author, and qutebrowser has limited userscript support. :(
It could probably be worked around by fetching the data externally, but I don't want to depend on hn-trustspark.com. It would be great if the profile updating tool could be published as well.
Anyway, hope it helps someone else :)
FWIW, after a quick review of the script, it looks safe to me. As long as you trust that the served profile data is correct, and don't mind leaking your IP to it.
Great idea and kudos to the author! We need more tools like this to help us deal with spam, and not just on HN, but everywhere.
[1]: https://gist.github.com/imiric/a5d1cd187e91eb0b725c4661250e5...
It's centralized for a few reasons though, first being that client-side API requests would be discourteous to the APIs (flood/ddos), and a whole new level of error handling would be required. Shared IPs, like those in a tech company building, would easily and quickly reach the API limits. So that's the reasoning, if you're curious.
One issue: New accounts with Zero submissions get a full green stack for "Submission Trust" (0 submissions in 0 days).
That comes across as three flat red lines and a singl tall green stack - not sure if that's the right message.
Submissions is an odd one ... those people that submit a story seen on the tech blogs once a day get bucketed together with habitual spammers of poor quality posts.
First is that I didn't want to make a plugin in the first place, I wanted to make a bookmarklet, but HN's CSP policy was too strict. So that was a bummer.
Second is that I have very mixed feelings about open source these days, and so open-sourcing feels less and less like the sensible default state. One of the sibling comments here discovered the alltrust.json and vibecoded around it, which is really a case in point about why open sourcing feels like I'd be leaving myself "open" to be domineered (not just by users, but by bots and companies as well).
Third is that the system/plugin is partly LLM-assisted itself (even though the code is minuscule), and I'm self conscious of being a slop-slinger. Or at least, pushing up repos with LLM code just feels, idk... lazy and asymmetrical (despite this plugin having clear utility, which I think it does).
But it's completely fair to say "oh look, a plugin about trust that's closed source, how hypocritical." I get that. If there's enough interest I'll open source it, sure.
Apologies for working around it and putting the code out there against your wishes. If you check my post history, you can see how opposed I am to these new tools, and "vibe coding" specifically. In my defense, I really didn't want to spend a lot of time on this, and LLMs do a decent job at this type of mechanical conversion. And I really don't judge anyone for using them mindfully, as you've clearly done in this case. The code didn't read like slop to me, if it's any consolation. :)
Besides, this "closed source" criticism is really a non-issue in this case considering it's a browser extension with clear JS, which anyone can inspect if they were really interested.
Cheers!
EDIT: I've deleted the gist. :)
Some more thoughts:
1. The website needs a description of what this project even is.
2. Open-source the code, especially since it's just for the community, I'd be vary of installing "some random" extension from the store.
3. The modal element in the iFrame on the website is cutoff on mobile widths using Firefox.
But it would be useful to know if I had up/down voted them significantly in the past.
A plugin like HN Comments Owl would be more useful IMHO.
(I'm not being facetious or accusatory, I'm genuinely interested learning how some of these high submission rates operate, since a lot look automated).
> what are you submission habits? Do you just submit links you find interesting often?
Pretty much. I will often email myself links I think might be interesting, then batch up submissions. I also like to highlight friends' posts.
> And does the karma rewards factor into your routine submissions?
It's always fun to get karma, but I like the fact I can call attention to someone or something that deserves it more.
I also wrote and talked about my thoughts on HN and how to be a good community member a bit more:
- in this blog post: https://www.mooreds.com/wordpress/archives/3530
- in this podcast: https://www.youtube.com/watch?v=VrObxK-J_3I
I think friend/foe and trust signals should come from a user's voting. So I think it should operate transparently to the user and there should be a default shade-out option of bad actors, but with the option to view. So it's set-and-forget. On the backend - could you make it so if you've never visited HN before, and you install the plugin, the experience changes accordingly?
I had the idea to setup a forum somewhere else - this sort of functionality would come in handy as part of normal operation.
Found one: https://web.archive.org/web/20260328080015/https://hn-trusts...
You have an overall 7 out of 8.
Is there a way to do this from a bookmarklet?
https://en.wikipedia.org/wiki/Trust_signals
https://en.wikipedia.org/wiki/Sparkline