New trust algo is looking pretty good. We basically use a statistical model to get a pretty good guess about the proportion a user's upvotes will be like another user's upvotes even from small sample size of upvotes.
There are a surprising number of folks who upvote well but never post, and many who post well but upvote poorly, so hopefully this will help good content surface faster/stronger and keep bad content down. It, too, is just an iteration but an iteration seemingly in the right direction.
I still need to figure out how we'll be adjusting daily rewards. It currently under rewards folks who are here frequently/contributing meaningfully and over rewards folks who are here infrequently and contribute weakly.
We are considering moving away from users getting rewarded for any past content, and only rewarding users based on the last 24 hours of content/revenue.
Because it force people to visit the page everyday...
I mean, sometimes a good news can be viewed and commented for more than 24h
I mean I think is a good idea to reduce the time, but 24h seems kind of rushy...
Is just an idea btw :)
You don’t need to visit everyday to get your rewards. You just can’t earn rewards tomorrow if you don’t contribute today. We can adjust it as time goes on … just trying to see what we can do to incentivize better content.
Right now it rewards a long tail of mediocre content and I think that’s bad for most users.
When you mention "upvote well", are you taking about the number of sats donated to a particular post, or the number of upvote interactions (independent of sat value) across a number of individual postings?
I mean, upvote like other users. E.g. to create a trust relationship between you and me, I look at everything you've upvoted (n, the number of trials) and see how many times I upvoted the same thing (s, successful trials). I then have a sample proportion of trust s/n.
Then I construct a binomial proportion interval with a 95% confidence level and take the lower bound. This is how much I trust you ... it works out nicely, because as you upvote more things, I get closer and closer to how much I actually trust your upvotes - yet even if you haven't upvoted many things I can make a reasonable (yet pessimistic) guess as to how much I should trust you.
Did the layout for long posts that would end up multi-line get changed (to truncate the display of the URL ?) Or has it been that way, and I just now am noticing it?
I should note that none of this is live yet. I have to decide whether I want to ship as is or also revamp rewards in the same release and pay down some related tech debt.
Lots of angry tweets about bitcoin maximalism right now.
Personally I think bitcoin is desperate for more development on the privacy side, a mass adoption right now is horrible for everyone, except the compliance NgU crew.
The people who treated peer 2 peer electronic cash as a casino is getting washed out. I think it can go as low as $3,000 and thats going to be very painful for everyone. It would take some sort of peer 2 peer rennaisance to go up again tho with many users coming to the space. And why not? Its really cheap at that point. The question is who can survive a dip to $3k. But lets see.
Also i obvously dont know whats going to happen :)
Mentioned this on Yesterday's daily thread that Wednesday, June 29th SN surpassed its previous all-time-high for daily uniques, but that was before the day-end total: 1,544!
The prior high (Friday of last week) saw 1,312 uniques, so this ATH was 17% higher than the previous!
SN Item #1: Jun 11, 2021
SN Item #5,000: Nov 15, 2021 (151 days since Jun 11, 2021) 33 / day
SN Item #10,000: Feb 3, 2022 (80 days since Nov 15, 2021) 62 / day
SN Item #15,000: Mar 17, 2022 (42 days since Feb 3, 2022) 119 / day
SN Item #20,000: Apri 14, 2022 (28 days since Mar 17, 2022) 178 / day
SN Item #25,000: May 3, 2022 ( 19 days since Apr 14, 2022) 263 / day
SN Item #30,000: May 20, 2022 (17 days since Mar 3, 2022) 294 / day
SN Item #35,000: Jun 8, 2022 (19 days since Mar 3, 2022) 263 / day
SN Item #40,000: Jun 30, 2022 (22 days since Mar 3, 2022) 227 / day
There are multiple implementations in different programming languages
I don't think multiple implementations impact centralization. There are dozens of reddit clients. Is reddit decentralized?
You can run the node on your phone ... and there's more LN nodes
Isn't that a custodial lightning wallet and not a node? Essentially using someone else's lightning node.
You can make transactions with your peer without anyone ever knowing about it
This is only true if you have a direct channel with them. Otherwise, all you have is plausible deniability. Besides, privacy is not decentralization. Using cash is also private, but that monetary system is also centralized.
There are many types of decentralization... e.g. single implementation is a centralized attack vector (manipulating devs, manipulating the company backing that implementation, etc). Having multiple different centers is helpful against some of those attacks.
What are you using as a definition of decentralization?
I quite like this definition of Trust Models (yeah... I know...)
"if the 100 routers with the highest simulated traffic increased their transaction fees to close to on-chain fees, the majority of payment sources would have no less expensive option to route their payments."
"90% of the nodes only correspond to 10% of the cumulative betweenness of all nodes. Consequently, this indicates an extraordinarily high network centralization, where 90% of the shortest paths in the network lead through only a few highly centralized nodes"
Removing shafts from a motor yacht while another team takes off the rudders. Once we take off the running gear others will move the boat into a building for several months of upgrades and maintenance I will not perform myself.
Thursdays I also have an apprenticeship class after work. Long day ahead of me! Best to everyone else!
n
, the number of trials) and see how many times I upvoted the same thing (s
, successful trials). I then have a sample proportion of trusts/n
.