Oddbean

Thinking about this some more — I think content curation can work, we just need to think about it differently. So with ontolo's model, you have (random) content, and you ask users to label it. The assumption is that the content being presented is worth labeling (or even looking at) in the first place. If we reverse the model, and ask people what they want to curate first, we can then use automation to narrow down content suggestions to then go into that bucket. This seems better because the curator has expressed interest in the subject matter, and has a much smaller pool of content to sift through. Instead of humans training AIs, AI could queue up higher value work for humans. The interface could work something like this: - What are you interested in? [categories/subcategories/custom label] - What scope do you want to use? [follows|network|global|relay] - Use heuristics, search terms, topics, and AI to present content to label. So for example, if they're curating pictures, don't show notes without a url in them.

I really like the idea of being able to organize content in to buckets or streams. Then you could subscribe to that stream, say I want the memes tagged by @hola_si_bueno_chau here instead of having them share them with me on IG. ;-D

Here is the current flow, and a proposed flow for ontolo. Note it is not specified which is a human and/or agent task. https://image.nostr.build/fbead607a3ce2f0a801eda114865f6eab713b95547144577d0d32bf8b26bb861.png https://image.nostr.build/55db5e765dccbab69febae452731e85e53905e3b50228f4171a9818ac74b0aaf.png

That is to say, content curation is arguably more valuable than content creation. It's less exciting, so people aren't as likely to do it for free, creating a supply shortage, increasing the possible market value. This of course assumes humans are needed for content curation at all. But someone could potentially make a living curating e.g. foot pictures more easily than by producing them.

Had kicked the idea around of just allowing free form entry of tags/labels but opted for something more structured to try and make it more universally useful. Maybe free form should just be an option. 🤔

A user doesn't have to click through multiple categories. There are various implementations, but here is one of the designs. I believe, we've already discussed it during a hackathon. Each event has only 1 category. Let's say, Alice labeled her event as 'memes'. When you see her event on the timeline, you will also see that it belongs to the 'memes' category. You can then adjust e.g. with a horizontal slider the amount of 'memes'-related events you want to see from Alice. By default, all categories for all authors will be at 100%, but if you want to reduce the amount of 'personal'-related and 'finance'-related events from Alice, you will change the slider for such events to e.g. 20%, while keeping all other categories, including your favorite 'memes'-related events at 100%. For better UX, clients can implement 'show more of similar content' and 'show less of similar content' buttons right near the events. The button to show more of similar content should be hidden if the category is already at 100%. Filtering can be done a client-level to reduce the amount of additional NIPs. We can also use other values/labels for the category slider like 'hot', 'rising', 'all'. Now the question is how to filter out which 20% or 'hot' events from Alice should appear on your feed? On Reddit-like decentralized social media platforms like DegenRocket it's very easy to implement such filtering because each category (e.g., DeFi, Privacy, All) is essentially a subreddit, which can be filtered by the amount of interactions (likes, dislikes, etc.) the event/post has. Voting manipulations though Sybil attacks can be mitigated by implementing whitelists or by token-gating your instance. On Twitter-like decentralized social media platforms like Nostr this can be implemented in a similar manner, but interactions will only count if they were submitted by users you follow in order to protect from Sybil attacks.