Ask HN: You are given 100M to launch a new Reddit competitor
what would you do?
Rearchitect the app to point it at the new site instead of reddit.
Model it after Wikipedia's monetization. Never advertise, maybe some monetization through reddit gold style status things you can buy but mostly keep the lights on by begging for money a couple times a year.
Make it nonprofit like Wikimedia foundation.
Keep reddit's structure of subdirectories. Replace the UI with what the app looks like.
Keep costs low initially by having videos and images hosted off site, text only.
Hard to say if this would make enough money to keep the lights on but would be interesting to see it attempted.
Replicate the api and link structures.
If legal, copy the existing content.
Give existing users a migration path.
Of course, you wouldn't be importing the info, the users would be doing it themselves.
I think this would make moderation worse. As a volunteer function, people do it for love of their community. As a job, there’s no competition and even more shenanigans for controlling large communities.
I think the direction should be toward making moderation take less and less time so more people are willing to volunteer.
since this keeps being passed down as fact.
while it's clear that the apollo guy is a bit awkward, he did clarify in the recorded conversation, that this was intended as a joke. [0]
I read it as he was trying to point out sarcastically how ridiculous the numbers were.
[0] https://gist.github.com/christianselig/fda7e8bc5a25aec9824f9...
There's a lot more to it than just launching a competitor platform.
Using NNTP would be a bit of a step backwards in some ways, however it would also allow distributed operation, and interoperability from a whole bunch of clients.
Honestly though I'd probably not really be too interested in duplicating Reddit. From users unaccustomed to paying, dealing with porn, dealing with abuse, a hard path to monetization, it just seems like a cute niche that wouldn't be worth investing in.
Which ways do you have in mind?
There's also a lack of formatting, but perhaps you could just not show posts without a header containing "text/markdown", or similar, to work around that.
Binaries would be grossly inefficient too, if you resorted to multipart-posts with encoded bodies. Again that could probably be worked around - either use single-post parts only, or ignore them entirely as you preferred.
What I'd probably do is spend a few million to build and do some advertising on a direct Reddit clone to see how that works, but then look to fund at least 5 promising social media startups with different approaches to communicating. Maybe something more people focused, more video focused, more audio focused, and other things.
Reddit isn't organized particularly well and I'd look for a smaller team trying something different. People don't always know what they want until they see it. Maybe one of those new approaches is really sticky and can be integrated into the Reddit clone?
Relaunch the Apollo app with the above strategy. Next, build a website to compete with Reddit.com and to get SEO.
Go from there.
I assume much of the moderation work is unpaid. It would be good to figure out a monetization strategy that doesn't involve advertising at all to pay the bills, and possibly allow moderators to have a strategy of their own that works to pay their bills.
Distributing it as open source would be a good idea, but it has to be built on a secure foundation, and Linux ain't it. I'd start with Genode, and leverage capability based security to allow far more robust management of information flows.
Invest 50M to cover all operating costs indefinitely using interest
Spend 10M in advertising and porting content
Sit on 10M as an emergency blanket
Invest 20M over time in startups and organizations protecting the internet
2) scrape the original reddit
3) using the data, fine tune a LLM to reproduce the type of content posted on the top subreddits (fake it till you make it)
4) launch multiple agents using the LLM to generate the content and interact with each other, in order to make the site feel alive
5) put ads for the Reddit competitor on HN, Twitter, popular forums and Reddit.
6) eventually start to turn off the agents.
1. Users can add links with tags 2. Users can upvote links 3. Users can curate their own set of links 4. Users can subscribe to other's links 5. Search can search on a combination of links, link content and tags
A modernized Delicious but crowd source relevancy without the need for moderation. Site becomes go to place for search in the same way I often get better results searching HN than Google for specific tech related items.
There are lots of former sysadmins from those sites. Hire them.
Bring in corporate sponsors. Sony might run the DSLR forum.
Have ChatGPT answer questions nobody else answers after x hours.
Profit.
1. Subreddits exist, but only as a way to group content. Creator of subreddit can determine things like css layout and whatever, but cannot set rules on posts, or ban users or delete posts.
2. Upvotes/downvotes on any content can be done once and is irreversible. Any up or down or total counts are hidden. Content (posts and comments) are scored by total number of upvotes, or total votes with sum being close to zero, weighted by the user scores (see below). Option to sort content chronologically as well.
3. Ability to add a special context comment to posts or other comment with option of (supporting evidence, opposing evidence, or unappropriated content), which can be voted on. Context comment that are upvoted or downvoted past 90% with a certain total number of votes, with their status moved from pending to accepted or denied. Context comments themselves can have context comments that work the same way, that can affect the status of the original context. Admins get notified of unappropriated content.
4. User content score and trust score, visible to everyone.
- Content score would represent posting good content, i.e something that gets a lot of upvotes, or a large number of votes with sum close to zero, just like in 1, or posting good comments under those posts). Initial votes on posts or contents with low counts (invisible to the user of course) will count higher towards this score.
-Trust score would depend on the user posting context comments or voting on context comments. Posting or voting in the correct way would increase score, posting context that is downvoted or voting in the opposite way of majority on context decreases score.
I would then use a large portion of funding to essentially pay people for content on a weekly basis, with payouts depending on the product of user engagement with some weighted sum of the content+trust score, and as a fraction of the total pool of money to be paid out.
The overall goal would be to do 2 things.
First, get people to auto scrape the web for good content and post on your app, giving you traffic.
Secondly, create a reliable place for people to read and interact without any bias. The monetary incentive combined with scoring will basically auto mod the website, since real money will be on the line and people would be WAY less likely to troll and shit post and think more about their actions to essentially bet on the most likely outcome in terms of posting stuff and voting, going off what they see alone (since scores will be hidden). Even more so, people would be competing against each other to be good citizens of the website to gain a larger chunk of the payout.
Once you have those 2 in place, you have created something that not only people use, but also something that effectively self moderates without having to remove or ban posts. You may have to do some tweaking on weighting algorithms, but overall most of it should technically work.
You can then take your app any number of ways. You can likely decrease payments for comments to zero, and people will still use it just like they would with Reddit currently, because they will have nowhere else to migrate until someone comes along and offers them money to do so. The user score will serve as an incentive enough to keep interacting, since high user score would take time to attain and thus have value.
The more interesting thing is that there is likely a way you can do decentralized content torrent style, where every leecher/seed is a user. Because every one of those has a score associated with it, it serves as a decentralized way to only serve actual content, preventing malicious actors from serving fake content.