Show HN: Grasswire – an Internet newsroom curated and fact-checked by everyone
grasswire.com
grasswire.com
Personal preference of course, but for instance after clicking Russia-Ukraine conflict it doesn't tell me in a quick glance what the story is, at the left there is a bunch of meaningless pictures from people I don't (or hardly) know, at the right some tweets in what seems at least 3 different font sizes, with the green imo making it harder to read,and then followed by a whole lot of videad with no title. Overall, an overload of colours and images. For me at least.
[1] http://pressthink.org/2014/01/keep-me-informed-parsing-the-l...
There are way too many low quality (twitter?) messages. Personally, I would not spend any of my time refuting and upvoting them. People have only 24 hour in a day, you need to provide better links, so people not only filter the data, but find new interesting stuff.
There's not really any good way to say "show me only good Twitter stuff related to this content." The top of the page should show you the best stuff, and the live streams will be for more hardcore users, kind of like the people who scan through the "new" page on HN. But we are definitely trying to find way to make that better if you have any suggestions.
Also, part of our thesis as to how that works is that information can be "important" even if it's false - it's important to mark it out as false, since it's going to be spreading regardless. So in theory the two actions should be separate, and the incorrect information should be marked as such. Could be wrong on that - we'll see.
In practice you're absolutely right. Will get on that.
Yet my wife was furious with me the first time she used grasswire for those same reasons. "I don't want to see that" is a reasonable request. So we have to find a balance there - we have to warn but not censor. That's tricky when things are live and we don't technically control what others are posting - it may end up as a blanket "stuff on this site might be graphic."
- The videos in the live stream are all really unappealing (not the content but the way they're presented), there should really be titles or comments or something on them. - The tweets move too quickly and the flashing is annoying, I have to keep scrolling to read them properly - The content above the black bar also needs some context besides just being pictures
- Hitting the ESC key I expect to close any open models, but instead I get redirected to the front page - Tweets with ampersands in them show up as &
- The videos under the bar seem to have consistently low resolution thumbnails, maybe try making them a bit smaller (to a resolution where they are meant to be in)?
- Many of the tweets aren't in English so they don't mean much to me, maybe detect languages and remove ones that aren't my native language / translate them?
- Just generally I prefer text over images or video for anything news related, and right now it seems like an overflow of images. e.g., Pinterest is almost all pictures but their 'main' content literally is whatever is in the picture (usually), whereas your 'main' content is the news, so I want to be able to just glance at a page for a news item and know without having to interact much further what is going on, similar to reddit.com/r/news.
- Also some other minor UI things that I noticed, the magnifying glass just seems strange, I guess it is a way to signal that this isn't a normal tweet but something I'm meant to pick out and observe further, but I might try doing that in a different way because it is just unconventional.
- It would be great to have a "score", like how many people have upvoted, how many confirmed and what their reference was, how many refuted and why; some level of discussion to go with it, etc.
Can't signup without clicking Login (slightly confusing to non technical users perhaps?)
The block layout gives no visual affordance to what is new / most active / has the most people verifying data etc. I can't see where the action is happening at a glance.
The text inside the boxes is scaled oddly and is actually difficult to read I find for the smaller boxes. I have to zoom in just to find out what some things are.
Videos have no titles. I have no idea what they're about without opening every single one.
I can't see who has confirmed the items (maybe it's somewhere?) seeing as you can sign in, being able to see WHO has confirmed a lead lets me check what else they have confirmed to help further solidly their authenticity (and perhaps expose bias).
Have you planned for people manipulating the site to mass confirm fake reports and bury legitimate data? Seems like an easy site to game with a few twitter profiles and a few accounts on your site.
1. You're absolutely right that we need to do a better job of giving a sense of where the action is happening. 2. The text issues in the boxes are definitely a problem and we have some tinkering to do on that front. 3. As far as videos go, we will add the titles. In general, the videos are not a great experience. It's difficult to find relevant videos using the YouTube API. We are considering not using YouTube directly, but instead surfacing videos that are mentioned in tweets that are deemed relevant; they usually seem to be more relevant and raw. 4. Being able to see who has confirmed/refuted what is definitely important and we need to make that experience better. For now, we don't even have profile pages that let you see what someone has done. We'll be implementing that in the next iteration. 5. The security/spam issue is a tough one and we've given some thought to it, but honestly, not enough. In some ways, we wanted to prepare for it, but we also didn't want to work on a problem that didn't need to be solved just yet. I am sure this will be a constant battle, but as we improve both our algorithms for filtering tweets and our API security layer, hopefully we can prevent bad actors from completely overruning the site.
Thanks again for your thoughts. Any other feedback is welcome and very much appreciated!
I'll keep an eye on this.
This service is not exactly it, but comes close to the idea...
Also: http://wiki.lesswrong.com/wiki/Debate_tools
If you find a good one let me know.
Very interesting concept though. You just need that critical mass of fact checkers.
I figure you're doing some live updating to the code right now, but just a heads up, I get two errors and a blank page atm:
XMLHttpRequest cannot load http://api.grasswire.net/v1/newsfeeds. No 'Access-Control-Allow-Origin' header is present on the requested resource. Origin 'http://grasswire.com' is therefore not allowed access. (index):1
TypeError: Cannot read property 'left' of undefined
at g (http://grasswire.com/scripts/8ca0d719.vendor.js:12:1650)
at Object.fn (http://grasswire.com/scripts/8ca0d719.vendor.js:12:1242)
at i.$digest (http://grasswire.com/scripts/8ca0d719.vendor.js:5:17771)
at i.$apply (http://grasswire.com/scripts/8ca0d719.vendor.js:5:19070)
at HTMLAnchorElement.<anonymous> (http://grasswire.com/scripts/8ca0d719.vendor.js:6:28105)
at HTMLAnchorElement.kb.event.dispatch (http://grasswire.com/scripts/8ca0d719.vendor.js:2:23617)
at HTMLAnchorElement.r.handle (http://grasswire.com/scripts/8ca0d719.vendor.js:2:20329)
EDIT: disabled it and ghostery, saw no improvement.
In some countries, this kind of information could be a powerful tool of repression. The police knocks on your door and asks "You helped confirm Rumor X about our Dear Leader. Tell us everything you know about Rumor X!" (Of course such countries also tend to block domains like grasswire.com, but some might choose to keep it unblocked for a while and use it as a honeypot to pinpoint dissidents.)
The multimedia-heavy, AJAX-heavy interface also makes it a royal PITA to use the site over something like Tor.
I've successfully signed up and in, but the newsfeed hangs (blank page): http://api.grasswire.net/v1/newsfeeds
All javascript etc. is also loaded succesfully.
Any ideas?
Also with ng-cloak, the angular script must be loaded in the head of the html, or the CSS rule from the script applied manually in the stylesheet: https://docs.angularjs.org/api/ng/directive/ngCloak.
Feature request. Allow me to add a reference when I "Confirm" it
1. 'like" is better than 'upvote' because it is clear that i am evaluate whether i like(re: enjoy) something, but what metric am i using to [up]vote on something?
2. "Reference URL" describes specifically what you want rather than some random URL.
Also, the lack of animation when moving the mouse off of an item is a bit jarring compared to when the mouse onto an item.
Surely there is some good/productive news worth sharing?
TypeError: Cannot read property 'left' of undefined
page is empty
Here's a screenshot: https://drive.google.com/file/d/0B2XTGTLKMtvmOHpYTnEyMEdnaE0...
The old one was just a list of tweets from the Twitter search API with an arrow that allowed people to move them up and down or.
The new one is a functioning newsroom, we do some machine learning to bring in youtube videos/tweets that we’ve identified to be relevant to a specific story, allow people to up vote, confirm or refute (while providing a link), take that all into some stuff we do on the back end and spit out content sized according to its importance (as a function of time, relevance, votes and fact checks).