What Now? Solving Hive’s Discovery Problem Without Selling Our Souls (Eng/Esp)
I don't know exactly how one would go about improving this, but it's a known fact that content discovery on Hive's front ends isn’t great. That’s not to say we haven’t made progress—clearly, we have—but for someone jumping in and logging in for the first time, they’re likely to be confronted with a blank feed.
That being said, I'm not suggesting a Hive front end should begin collecting personal data—at least not in the same way Web2 platforms do—but something can be done, and I’d argue should be. From my limited experience (because I’m not the type to go out there hunting for the latest social media apps), it tends to work something like this:
A brand new account, once created, selects from a list of topics of interest. It could be something broad like music or sports, or something much more specific. Some platforms even let you type in the kind of content you’d like to see. From there, a personalization algorithm starts working to deliver a feed the user might actually enjoy.
Over time, through a mix of accepted and rejected content and clear documentation of engagement metrics, these algorithms fine-tune themselves. And this is where things start to get... icky.
In the Web2 world, the holy grail is engagement. No other metric matters as much. If a type of content generates more clicks, more comments, it gets pushed to the top—period. This has consequences, obviously. We humans are suckers for rage bait, and no one seriously paying attention can say that's a good thing. But here’s my question: is there a compromise we’re failing to see?
What I’m talking about here is essentially a layer—or layers—built on top of a “database.” I use air quotes deliberately, because a blockchain isn’t really a database in the purest sense. But you get my point. The more I think about it, the more I believe this would require a solid backend implementation—and yes, a degree of data centralization. That’s a phrase we’ve been trained to reject.
So here’s the challenge. For you, for me, for all of us—a community of dreamers and status-quo rejectors:
Can we come up with a way to decentralize discovery algorithms?
Can we build a front end that caters curated content to its users, keeping them engaged and enjoying themselves—without sacrificing the ideals that brought us here?
I’m fully aware this is the thing social media tycoons obsess over. How to make platforms more addicting. More pleasant to stay on. But they’re not thinking about balance. That’s why I used the word “addicted.” I think there’s a way to find balance, even if I can’t fully see the solution yet.
In case you’re wondering why this is on my mind, it’s simple: I’m working on a front end myself—an app to make Hive Snaps easy to use. And while it already answers the “Where’s the feed?” question pretty easily, the truth is, Snaps aren’t curated at all. You just get everything people are talking about. Which, frankly, might be a good thing, at least for now.
Still, I can’t count how many times this year alone I’ve onboarded someone to Hive, only to watch them stare at a blank feed and ask:
What now?
MenO
Spanish Translation
¿Y ahora qué? Repensando el descubrimiento de contenido en Hive
No sé con precisión cómo podríamos mejorar esto, pero es un hecho conocido que el descubrimiento de contenido en los front ends de Hive no es el mejor. No quiero decir que no hayamos avanzado, eso está claro, pero para alguien que entra por primera vez y se loguea, probablemente se encontrará con un feed en blanco.
Dicho eso, no estoy sugiriendo que un front end de Hive deba comenzar a recolectar datos personales—al menos no como lo hace Web2—pero algo se puede hacer, y diría que debería hacerse. Desde mi experiencia limitada (porque no soy de los que están probando nuevas apps sociales todo el tiempo), esto suele funcionar más o menos así:
Una cuenta nueva, al crearse, selecciona de una lista de temas de interés. Puede ser algo general como música o deportes, o algo más específico. Algunas plataformas incluso te permiten escribir qué tipo de contenido quieres ver. Luego, el algoritmo de personalización empieza a armar un feed que posiblemente te guste.
Con el tiempo, gracias al contenido que aceptas o rechazas, y al registro claro de métricas de interacción, estos algoritmos se ajustan. Y ahí es donde la cosa se pone un poco... turbia.
En el mundo Web2, lo importante es la interacción. No hay otro indicador que pese tanto. Si un tipo de contenido genera más clics, más comentarios, se impulsa al frente, simple y llanamente. Eso, por supuesto, trae consecuencias. Nosotros, humanos conflictivos, no podemos evitar caer en el rage bait, y nadie que esté realmente prestando atención puede decir que eso sea algo bueno. Pero me pregunto: ¿hay un punto medio que no estamos viendo?
Lo que estoy planteando es más bien una capa—o capas—sobre una “base de datos”. Uso comillas porque la blockchain no es exactamente eso en un sentido estricto. Pero entienden a lo que voy. Cuanto más pienso en esto, más creo que necesitaría una implementación robusta del backend—y sí, cierta centralización de datos. Lo admito: es una frase que estamos entrenados para rechazar.
Así que ahí va el desafío. Para ti, para mí, para nosotros como comunidad de soñadores y rebeldes del statu quo:
¿Podemos idear una forma de descentralizar los algoritmos de descubrimiento?
¿Podemos construir un front end versátil que ofrezca contenido curado a sus usuarios, manteniéndolos interesados y disfrutando de la experiencia?
Sé que esto es precisamente lo que obsesiona a los magnates de las redes sociales. Cómo hacer que sus plataformas sean más adictivas, más agradables de usar. Pero a ellos no les preocupa el equilibrio. Por eso usé la palabra “adicto”. Aun así, creo que podríamos encontrar un equilibrio, aunque admito que no tengo la solución del todo clara.
¿Y por qué estoy pensando en esto? Fácil: estoy desarrollando un front end propio, una app que hace que Hive Snaps sea fácil de usar. Y aunque ya responde bastante bien a la pregunta de “¿Dónde está el feed?”, la verdad es que Snaps no tiene curaduría. Simplemente te muestra todo lo que se está hablando. Algo que, seamos honestos, puede ser preferible, por ahora.
Pero ya he perdido la cuenta de cuántas veces este año he ayudado a alguien a unirse a Hive, solo para verlos mirando un feed en blanco y preguntar:
¿Y ahora qué?
MenO
I was puzzled by this post because I was thinking, this is already solved, you've done it on Snaps ... the feed with the newest posts ... why can't people choose from that and start to build their preferences through follows?
Oh yes, it works for Snaps somewhat... but, not at all on peakd or Hive.blog - the flagship interfaces to Hive.
I'm convinced I won't be onboarding people with the "traditional" front ends for a little while. Start them on Snaps, make them make friends, and then be like... "Hey, try this..."
Luckily for me, i have day to day convos with newbs. So I'm very much aware of their confusions. lol
Oh yes, it was overwhelming when I first joined! Snaps is a great starting place.
if you ask someone to select from a list of topics of interest, you are already collecting data from that user. I personally don't see a problem, as the user can choose to provide the data or not. Maybe save that info in the user metadata to create a feed.
At the same time, a user can choose communities of interest once they are onboarded, the onboarder should show how to do that, and the user can click to see the communities feed (link on peakd: https://peakd.com/created/my). This is kinda like picking the list of topics you suggested, and that feed is not blank anymore. It's just not the default feed in the frontend.
Oh.. i like this as part of the metadata... good thinking... hrmmm
For new people, I usually say to post a couple things that I can share and upvote, suggest a few fun people to follow, point out the communities, and then to follow and curate the active people who reply to my posts.
Find the active people, support them, be one, explore, learn and network.
Search definitely would be a valuable add though.
have you ever played around with snaps?
You know, I haven't!
I have seen the posts up to 'third time's a charm' and promised myself I would jump in and help out but haven't made the time yet.
I suppose I could search but how can I try?
if you are on peakd, you could click on snaps right on the top. Second button... you will see some of us there too, making short content. That said, you could also join my beta testing group. I got android and iOS. A dedicated app for short form content on Hive.
I could do both!
How do I get the iOS ap?
info here muh fren...
see you in discord
https://peakd.com/hive/@meno/hivesnaps-ios-beta-calling-testers
I think the best way for new users to get traction is to encourage comments on topics that interest them. Engage with other writers in a genuine way, of course, not just "nice post" filler.
Communities can help find such content, but not everyone posts in communities. Tags help, but not everyone tags their posts well. What if there were a way to easily see what a user curates over time?
What if we could open a tab and see, "so-and-so regularly manually upvotes users A, B, C, and D" as well? I assume an algorithm could check for auto votes and trail votes using behavior patterns if nothing else.
Not a bad idea... but it would to have to be an algo that builds a follow list based on manual curators, otherwise, circle jerks would get mixed in there too.
"does this user usually vote after the first 30 minutes a post is live?" Would winnow out a lot.
https://x.com/jewellery_all/status/1949195761977348132
#hive
One simple thing that could be done is for the newly-created account to follow their onboarder account.
And for sure, content discovery algorithms can work in decentralized ways. For example, it could start with suggesting topics that are similar to the interests of the onboarder account. But do so in a way that makes it clear this is just a taste of all the topics. So you can get a taste of it to start with but then you can explore more and more as time goes on.
And one note that could be worth keeping in mind. Traditional social media has a huge problem of echo-chambers. The more you follow a topic and a viewpoint, the more you are served that same thing. So it helps create extreme and difficult-to-change views. Focusing somewhat on algorithms can be good, but focusing too much on perfecting them might bring about the same problems.
This is the problem that I assume hashtags were meant to solve. If you query tags for things you are into then you should find some interesting people. Of course many just use tags to try and earn more tokens (as you have too). We could have tools to automate this discovery, but it needs to handle abuse. If a post has 30 tags then it can probably be ignored. The posts and users need to be getting some votes to show others care about them.
I'm not even sure that Trending is the best place to search. Communities may be a better bet if there are some for your interests. A new user should be presented with some options to find content they would like, not just left to themselves, but they should be able to opt out if they like.