Firstly, you need to get out of your own way. Most people start editing before they finish writing, which short-circuits the process of exploring ideas.

Getting feedback from other people is usually helpful but slow. And unless you have a professional editor, there are only so many requests you can comfortably make of friends and family.

AI completely changes the feedback landscape for writing. It is faster and has unlimited patience – not to mention being trained on the entirety of language. For the vast majority of people and circumstances, AI editing is superior to soliciting human feedback.

The challenge with AI editing – like getting words down initially – is not to short-circuit the exploration process. It is dangerously accessible to dump a paragraph into ChatGPT and ask “what do you think?” But seeking immediate feedback can crowd out the mental space needed to figure out what we are trying to say.

As with other deeply personal disciplines, the key to using AI in writing will be figuring out how to use it as a thought partner rather than an executor. We now have the tools to design infinitely better feedback loops, if we’re thoughtful about how we use them.

• Take care of my body: This definitely includes exercising more, but more broadly, I’d like to pay attention to how I’m feeling. For example, this year I realized sugar was causing me to crash hard, so I’ve cut it out as much as I can during the day (and save it for dessert).

  Exercise makes me feel great, so I need to discover which types of movement work best for me.

• Write more: After struggling to get back into the practice for a few years, I started First Draft Club as a way to force myself to write – and I’m now on a 20+ week streak of sharing something publicly.

  I want to keep pushing myself here. Quantity leads to quality, so I’m not upset about the consistency, but I do want to take some bigger swings. At the end of 2025, I hope to have a few essays I’m proud of.

• Read more: This past year wasn’t as good for reading. I spent too much time online, and most of the “reading” I did was short-form. In 2025, I’m hoping to carve out more time for deep reading.

  Since I tend to finish fiction books more consistently, I'll use novels to rekindle my joy of reading. For non-fiction topics, I'll focus on essays and articles, which better match my learning style than books with a lot of fluff.

• Engage with my local community: While I've built a strong online community, this year I’d like to deepen my real-world connections in New York City. This means connecting with both people and places – from neighborhoods to local institutions.

  I want to be a more active presence in my neighborhood – getting to know the baristas at my local coffee shop, showing up to weekly game nights at the bar on the corner, and becoming a "regular" in a few places.

  Beyond the neighborhood level, I want to build on my growing interest in city politics. While I’ve started paying more attention to what’s going on, this year I want to put "boots on the ground" and find ways to contribute to positive change in my city.

The motivation here is both practical and idealistic. For the visionary: this could be a meaningful way to combat the loneliness epidemic and help people build community, especially in relatively transient environments. For the operator: this a large population with distinct needs and strong potential network effects. Plus, there’s optionality with monetization down the line.

In 2022, more than 20 million U.S. households lived in apartment buildings with at least five units. About one third of that total lived in buildings with at least 50 units – and that population grew 8% from the previous year.

There is no platform that is designed specifically for the social needs of the urban tenant. The current "competitors" in this space – to the extent that they are used at all – are general purpose tools like WhatsApp or Facebook groups, or clunky forums run by building management.

A successful platform would make it easy and fun to connect with neighbors, focusing on use cases that provide value. Make it easy to host a shared Halloween party on the roof or organize a Super Bowl watch party. Empower residents to share resources like printers and vacuums or split bulk orders from Costco. Help neighbors find someone to walk their dog or even watch their kids for an evening.

I think the go-to-market would be relatively straightforward, especially compared to other "consumer social" ideas. Just as Tinder and Bumble targeted fraternities and sororities on college campuses, founders of a "Gen Z Nextdoor" would benefit from network effects and a repeatable playbook.

Target one building at a time, learn what works, and move quickly to the next building. Find natural community leaders, listen to their feedback, and build for them. Figure out what types of buildings make the most sense.

Eventually, once you have traction across multiple buildings, you can start working with the building management companies. Integrate the network with building infrastructure to create a more tenant-friendly living experience. When the network becomes popular and valuable enough, it becomes its own amenity – one that management companies will purchase to attract new tenants.

GPT stands for Generative Pre-trained Transformer.

The use of custom GPTs or tailored chatbots may be old news for some, but to me it was the first time I had a visceral experience of dynamically interacting with someone else’s knowledge.

The printing press was revolutionary because it allowed people to encode their knowledge and widely distribute it for others to engage with. However, in our modern society, books are faulty mediums for learning. They are static texts that do not adapt to our context or answer our questions directly.

ChatGPT and other large language models, on the other hand, are interactive but not specific. They are trained on a massive trove of text, not designed to facilitate learning within a specific domain.

In the past year, I have seen some chatbots trained on focused text, such as the work of Naval Ravikant or the writings of Lenny’s Newsletter (and friends). I am sure there are others. But I believe we are just scratching the surface with GPTs. Anyone who writes a book should have a complementary piece of software for readers to engage with – and I suppose in many cases, this software will replace the need for a book altogether.

Beyond chatbots, there is an emerging design space for encoding knowledge into interactive software. Lex, an AI-enabled writing processor, is developing a Prompt Builder & Library that serves as a marketplace for writers to leverage the expertise of others to inform their writing.

If the printing press – and later, the internet – unlocked “reach” for knowledge, then trained language models will unlock “engagement” for expertise.

I'm looking forward to exploring custom GPTs in the coming weeks. Please let me know if you have any favorites or recommendations!

Unfortunately, I also tied Angel User into my identity—it became my "personal channel" on Farcaster and the name of this Paragraph publication. This unintentionally created pressure to turn it into a fully developed pursuit. (I had been publishing screenshot essays to Farcaster instead of full posts here because the posts didn't fit the "Angel User" brand!)

Mark Fishman on Warpcast

Decided that a personal channel would be the best way to journal in public you're free free to follow along /angeluser

Farcaster

My first Farcaster project, Concept Testing, also bit off way more than I could chew.* So I learned some lessons when, a few months ago, I started my next project.

Although I originally had grand plans for First Draft Club, I was able to talk myself down into a small experiment: a 4-week commitment from 8 people to write something each week.

Over time, FDC has slowly grown and evolved. But I have made changes deliberately, adding new wrinkles only when I feel they are necessary.

It's been helpful to think of it as just a project – a mere subset of my identity. First Draft Club isn't a startup or a full-time pursuit.

This insight frames how I'm thinking about Angel User now. Instead of forcing myself to make it work, I'm letting myself build something organically.

The "MVP" for First Draft Club was relatively straightforward: I needed accountability to write consistently, so I recruited some friends and added a financial incentive.

What does that mean for Angel User?

Well, the first step of being an angel user is using and sharing helpful feedback on products. It's something I truly enjoy doing and also think I'm good at.

Mark Fishman on Warpcast

Being an angel user means using & supporting products that you want to exist – even if they aren’t the best products at that moment

Farcaster

Over the past year I have explored various ways of soliciting products to give feedback on – including cold outreach, pop-up services, and even a few livestreams.

The livestreams were the most rewarding. After sharing the recordings with the respective founders, I got strong positive feedback on the video format. Since then, I've embraced the practice of sending short (<5min) video recordings to founders after using their products.

Feedback videos are easy for me to do, and they provide concrete direction to builders about parts of their product that are confusing, buggy, or lacking.

Again, I think it makes sense to stick with an organic approach. Instead of requesting people to request feedback, I'll just naturally send videos to people whose products I've tried and enjoyed.

If those people value the feedback, they might ask for more, and from there, a working relationship could naturally develop.

It's similar to what Tom Critchlow did when he articulated his "dream clients" in a blog post. I don't know what my work will look like yet with Angel User, but it makes sense to build relationships with the builders whose work resonates with me.

It may not seem like much externally, but stripping down this internal expectation for myself is freeing. Angel User is now just a concept, not an identity of unfilled potential.

Rather than trying to figure out how to pay the rent with Angel User, I’m going to be focusing on making genuine connections and contributing valuable feedback. The rest will follow in time.

*Technically Concept Testing is still alive – we're still waiting on the devs to finish Cozycast. But we probably won't launch another Concept unless we radically simplify the structure of the project.

If you’d like to get feedback on your product, book some time on my Spotlight!

This post wasn’t sponsored by Spotlight, but I do like to write about products I'm enjoying – so it's possible your project could be featured in the future 🙂

Spotlight is a new product that serves as a “Farcaster Attention Network”. Each person on Farcaster is given one spotlight per day that can be purchased at a price that is programmatically determined by the historical demand for that spotlight.

Farcaster

MJC on Warpcast

/spotlight is live. Spotlight is the farcaster attention network. It’s a social app. It’s a creator monetization platform. It’s a media protocol. Spotlight is a new business model for the internet, designed to efficiently monetize the work you do and the attention you command. https://www.onspotlight.app It is *not* a friend-trading app. It is *not* a tipping protocol. This video introduces the project and demos the app. Some more details below 👇 https://www.onspotlight.app

When I first heard of Spotlight and got a demo from Michael, I was immediately intrigued. I really like the idea of using constraints for attention and monetization.

To illustrate with an example: I have considered starting a Hypersub for freelance services, but that creates an equation with multiple variables to solve for: the price, the number of people I could serve, the length of time I’d serve them for…

The genius of Spotlight is its simplicity. There’s one unit per day, and the price is determined by the demand. The business model is fixed; now you just figure out what to sell. Because Spotlight is set up as a daily offering, it constrains what you can offer. The challenge –and opportunity– is to figure out what your atomic unit of value is. (I'm trying 30-minute product feedback calls.)

The current priority for Spotlight is bootstrapping activity on the network.

The product already has important building blocks in place; everyone on Farcaster already has a profile and a wallet built in. This makes Spotlight “permissionless” – someone can book another person’s spotlight without them even knowing Spotlight exists.

However, I think this streamlined onboarding that actually presents an issue for Spotlight. For people to feel confident booking spotlights, they will want to know that someone is actually active on the platform. I don’t want to just send someone money without knowing if I’ll actually get what I bought.

While it is nice that someone’s profile already exists, it should be clear whether that person is active on Spotlight. One simple way to do this would be to show an icon that indicates that a person has “set up” their profile.

• Does the user have a description of their spotlight use case and booking policy?

• Have they opted into DCs for booking notifications (not currently a feature, but something I expect to be built)?

In the future, I could see something like reviews being part of Spotlight – but that’s a longer play. There are some small tweaks here that I think would increase consumer confidence and boost conversion.

Conversion is just one factor for demand; the other is discovery. Assuming Spotlight can onboard enough active “supply”, they will need to help users find the right spotlights to book.

One way to do this is with categories. There are some initial use cases that the team has identified, and I think there are ways to lean into these in a lightweight way without overly committing to building out specific workflows.

For example, allowing users to append a single tag to their profiles (ie. “Advertisement” or “Consulting”) would give Spotlight an opportunity to curate different spotlights across the app. This is similar to how Bountycaster scaled their discovery on their marketplace. Right now, the search/discover page on the Spotlight app has a list of “trending” users – bringing categories here would help with discovery.

Finally – and I know the team is already working on this – the booking flow should be simplified.

Right now, booking a spotlight requires publicly sharing an image, message and a link. This is used to generate content for the feed, in addition to serving as a communication channel for the booker to send information to the receiver. However, reducing friction for consumers is key; only the information necessary for booking should be required.

A @spotlightbot can certainly exist as a way to drum up attention on Farcaster feeds, but this shouldn’t be the responsibility of a booker. (I have had people tell me they didn’t want to book my spotlight because they didn’t want to publicly share what they were working on. Of course, they didn’t need to share that specific information on the public booking flow – but I understood why asking for a message & image during the booking process would be confusing.)

My recommendation is the booking flow should be treated as a transaction split into a few simple steps:

1. Confirm spotlight information (category & service offered, booking policy)

2. Choose a date and confirm transaction

3. Send requested information to spotlight owner

I have shared some of this information with Michael already, but I wanted to do a writeup partially to share my excitement about Spotlight. As an angel user, I want to support products I want to see in the world.

But also, I'm trying to scale up my Angel User practice, and I want to demonstrate the type of feedback I like giving to builders.

If you'd like to get your product on my radar, you know how to reach me 🙂

Farcaster

Mark Fishman on Warpcast

Being an angel user means using & supporting products that you want to exist – even if they aren’t the best products at that moment

Normally, when there are no external motivators, I won’t hold myself accountable. A “deadline”, a calendar block, or even a reward won’t work for getting myself to do something that I don’t exactly want to do in the moment. (Hello /adhd fam!)

There are many reasons that people may struggle to write. Impostor syndrome is probably most prevalent, but externalizing thoughts can be a scary process even if you’re not going to share with anyone but yourself.

So, why am I actually publishing this time?

I created some external motivators.

Farcaster

Mark Fishman on Warpcast

First Draft Club v1: a writing accountability group with 4-week commitment - Buy-in is $20 via /daimo - Share new writing in /firstdraft before Monday 9am ET each week - Each week you ship, you get your $5 back - Groupchat for mutual support Have a few slots left, DC me if you want to join!

My previous life as a behavioral scientist taught me that there are plenty of ways to incentivize behavior. Whether the pressures are financial, social, or psychological, I knew it was possible to get myself to write more. (Why I love writing is another post.)

I had been toying with the idea of a writing accountability group for a few months. After some discussions with friends and group chats, I realized it made sense to ship an “MVP” that wasn’t too much of a technical lift. Instead of obsessing over the idealized version, I just decided to “ship” the bare minimum to start – just like I’m doing with this post.

So here we are! Thank you to the OG members of First Draft Club for taking a chance on this and shipping this week.

I would love if this inspired you to write. But I also understand if you need a little more support :)

Feel free to get in touch if you want to work on any of these! You don't need our permission, of course, but we're always excited to help with scoping projects.

Okay, sorry for the self-promotion, but...

In case you haven’t seen it yet, Cozycast is a fully-scoped concept, complete with a $3400 bounty raised by 73 contributors / beta testers.

From our original post:

Cozycast is a way to filter the firehose of Farcaster into feeds that matter to you.

We envision a simple app that is designed to let users stay focused on the people, channels, and communities of their choice – free from additional noise.

Using a mix of the Farcaster social graph and onchain data, Cozycast users will be able to create feeds to their own liking.

Check out the full Cozycast spec here!

There’s so much more to someone than what you can learn from their Warpcast profile.

Even if you just use publicly available data, it’s possible to give a more “personal” look at someone’s account.

If done well, people could use this tool to both research new accounts and to promote their own profiles.

One signal that there is demand for this: a number of people asking for the ability to pin casts to their profile.

We want to share more about ourselves!

Some info I’d love to see on a profile (ordered roughly by the amount of “signal” they provide, not necessarily how easy it’d be to build):

1. Channels the person hosts

2. Top channels they post in

3. People they engage most with

4. Mutual follows / followers (like Warpcast does now)

5. Top casts from past [30] days

6. Root cast / reply ratio

I envision the profiles existing for every account automatically on Farcaster (hooray permissionless data!) However, there is an opportunity to let individuals claim their own profiles and further curate the information displayed.

Instead of “top channels Mark posts in”, I could list “some of my favorite channels”. An unclaimed profile would automatically show the top accounts that person engages with, but a claimed profile could tweak the list to show “some of my favorite follows”.

When a user claims a profile (maybe supported by an NFT purchase?), they could also add more personal information, including pinned casts, links, or even photos.

Search on Farcaster is due for an upgrade. Do you know how long it took me to find this cast?

Farcaster

Mark Fishman on Warpcast

Based on the replies here, my top suggestion would be to enhance search Better search functionality 👇 Find relevant conversations 🔁 Connect with like-minded people https://warpcast.com/dwr.eth/0xbc72d0

The original Searchcaster.xyz is legendary, but the "advanced" version is simply searching by account. I know Greg is no longer maintaining the site, and I've experienced some reliability issues fetching newer casts.

Here are some things I’d expect from a modern search tool (again, roughly in order of importance):

1. Filter by user, channel, and date range

2. Granular text search (boolean operators and phrase matching)

3. Filter by people I follow

4. Filter by links or media

5. Root cast vs. replies

6. Filter/sort by number of replies or likes

7. Filter by account stats: power badge, FID < x, etc.

I would happily pay $5/month for something that does half of this 🙂

I’ve seen a few asks for a “StumbleUpon for Farcaster channels” in the /someone-build channel. There are so many new channels being created that it’s nearly impossible to stay up to date.

Here’s how I imagine a “channel surfer” would work as a website. Each channel has a page with (some combination of) the following information:

• High-level info

  • Channel description

  • Host info

  • # of followers

    • if signed in: # of followers you follow

  • (If possible: surface channel pass cost or automod rules)

• Activity level

  • # of root casts in past 24h/7d/30d

  • # of unique posters in past 24h/7d/30d

  • # of replies per cast

  • Follower growth in past 24h/7d/30d

• Content

  • Pinned casts

  • Trending casts

  • Media

    1. Top link domains shared

    2. % of root casts with links

    3. % of root casts with images / videos

    4. % of root casts with frames

What else would you want to see about a channel that gives you a good overview?

Of course, each page (including the home page) would have a "Jump to random channel" button.

I don’t know if there is a “right” way to do this, but it would be neat to see related channels when on a specific channel page.

This would take the channel surfer from being a pure randomizer to more of a guided discovery tool.

I hope this inspires some ideas for those who are building over the next few weeks. We're looking forward to seeing what everyone creates!

New to Concept Testing? Read our announcement post.

Need a refresher on Cozycast? Check out the previous post.

We almost doubled our initial target on our crowdfund for Cozycast!

Farcaster

Mark Fishman on Warpcast

Absolutely ecstatic about how our first /concept-testing crowdfund went Thank you to everyone who contributed to the Cozycast campaign – more updates soon! Next step: putting funds into a bounty to get Cozycast built 🚀 https://i.imgur.com/Vu8ZmG2.png

One step closer to bringing custom feeds to Farcaster.

Now, on to the fun part!

If you contributed to the crowdfund (or previously minted a Concept Testing post here on Paragraph), you can join the group chat for beta testing.

Farcaster

Mark Fishman on Warpcast

If you contributed to the Cozycast crowdfund or minted a Paragraph post from /concept-testing, you can beta test the Cozycast MVP! We airdropped you an NFT that should let you join this group chat 😃 https://warpcast.com/~/group/Z4cFp1rK_61gf2B_RfJ3mA

1. We post a bounty with details on the process and development spec (what you're reading now)

2. Any devs interested in building Cozycast can reply to the bounty to apply. You must link to your Github with a hosted, early version showing that you "understand the assignment" and are capable of delivering the spec (see below).

3. Once approved, we'll mark the bounty as "in progress" and add the dev to the group chat with beta testers.

   

   "In progress" means someone is working on it, but anyone else can still post a GitHub early version to apply and deliver better / faster.

4. Devs post updates of their Cozycast version in the group chat and receive feedback from testers.

5. Once the version fully meets specifications and beta testers like the user experience, the bounty is marked complete and paid out.

1. A web & mobile app (Android + iOS)

2. A default "Following" feed

   • All casts from people you follow, in the channels you follow

3. The ability to create custom feeds with additional variables:

   • Casts restricted to users with FID above/below custom #

   • Casts restricted to users with a specific NFT in a connected wallet

   • Casts only in specific channels

     • This can be done either by muting (subtracting channels from initial following) or curating (adding channels from zero)

   • Casts from all people vs. people you follow

4. The ability to open casts in Warpcast

   • Cozycast is "read-only" by design, so in order to cast or interact with frames, it will need to be part of an app with signers.

• Your code is expected to be typesafe using Typescript and have working builds for Web, iOS (expo go or Testflight) and Android published to the beta testers.

• Functionally you need to fulfill the requirements laid out above, while the user interface will be judged and feedback for improvement provided by the beta tester group chat as they contributed to the bounty and make this all possible by putting in their money and time.

• The actual code has to handle errors, be deployable to production straight away with documentation on how to deploy available.

• If there is no network connection or other errors, notifications are shown. Loading animations are used when the network connection isn’t instant on interactions.

• Feeds are infinitely scrollable so if you hit the end it loads more, when available.

• Interactions are animated and the UI as clean as possible. Remove or use context menus instead of clutter.

  

  When in doubt: don’t overthink the delivery, ask beta testers for feedback and iterate.

Further development notes from Samuel

There is general technology needed to build any kind of Farcaster thing and also specifics for building the proposed MVP in terms of a read client.

All of them can be accomplished via Expo React Native in Typescript. Expo provides the rails to write Typescript code once and deploy to native iOS and Android as well as the Web for browser and a server to host potential API endpoints you may need.

In general all casts (content published to Farcaster and therefore also Warpcast) is present on so called Hubs. A Hub is a computer running software to receive and propagate Farcaster posts as well as hold information on user profiles (FIDs).

Anyone can spin up their own hub and read/write to the network.

For speed many people like to use the traditional database PostgreSQL and then synchronize with a hub.

The database is used for quick read/writes that are known to scale in traditional computing environments. To synchronize data from the PostgreSQL to the Hub and from the Hub to the PostgreSQL database there is a NodeJS based replicator available.

With that for reading you could leverage any regular database read mechanisms you already know.

Also for reading there are React Hooks already built by Dylan working with Neynars APIs so you don’t need to run your own replicator/hub and there are typescript - expo client implementations you can look at and potentially just adjust their feed.

• Quikcast by Horsefacts

• Opencast by stephancill

• Litecast by Dylan

Where for APIs there is also Neynar Developer APIs and an awesome hub to use from Pinata.

Currently as of April 2024 there is work being done to make the React Hooks compatible with straight requests to Hubs or the PostgreSQL replicator so you are free to choose managed solutions or run the infrastructure yourself, though the easiest solution to get setup right away is to use Neynar APIs.

Relevant Neynar APIs:

• Get latest casts and apply some basic filter: https://docs.neynar.com/reference/feed

• Get a feed to show casts from people you are following: https://docs.neynar.com/reference/feed-following

1 - Set your target on the clear default filter you apply and what potential options you want to explore with users (e.g. FID below 20.000 or only nouns token holders, …)

2 - Use farcasterkit.com React Hooks to get Latest casts for a reverse chronological feed and filter what the users needs in his feed by filtering for if user follows and filters don’t block

3 - In the basic user interface link out to Warpcast so if homefeed shows a cast one can act on that cast in Warpcast and view replies there too

4 - Step away from just one user used in testing and add authentication (sign in with Warpcast via Farcaster Auth Kit)

5 - Clean the user interface so the app is a joy to use

6 - Add reply viewing where with interaction user could see the full conversation with filters applied without having to go to Warpcast

7 - Add channel support so user can browse channels based on a list of all of them - here only the filter would apply and cast from people the user does not follow would be shown too.

8 - Add muting channels. If a channel is muted no casts from that channel will be shown no matter if it matches the filter or not.

9 - Add favorite channels where users can favorite (follow) a channel and pin that to their user interface or a sidebar

10 - Add a feed where one can select multiple channels and view a curated list of channels together. E.g. I select /spirituality and /philosophy so now I am viewing casts from both + the filter I applied

11 - Clean the user interface so the app is a joy to use

12 - Add Frame support (could use Expocaster to add them into your Expo App)

If you want to go further: 

13 - Add Signer management and write (react, respond, follow from feed) - Sign in with Neynar makes it easy

14 - Add user profiles page when clicking a users name

For more about Cozycast, keep reading.

For more information about the meta-project and how Concept Testing works: read our announcement post here.

Earlier this month, we announced our intention to work on a project called Cozycast.

You can find the full post with our motivation and explanation here, but I wrote the following:

Cozycast is a way to filter the firehose of Farcaster into feeds that matter to you.

We envision a simple app that is designed to let users stay focused on the people, channels, and communities of their choice – free from additional noise.

Using a mix of the Farcaster social graph and onchain data, Cozycast users will be able to create feeds to their own liking.

After gathering feedback on this idea, we decided to move forward and scoped out what an MVP of Cozycast would look like.

So: what exactly is Cozycast?

Well, that depends how much we crowdfund for the idea.

More funding translates into a larger bounty, which can entice developers into more complex projects.

However, as with all Concept Testing projects, Cozycast will be open-source software. When finished, it will be open for anyone to use for their own projects.

Although we will not host Cozycast as a standalone site or app, we are hopeful that builders of existing and future clients will implement the code into their own products.

Farcaster

Mark Fishman on Warpcast

The proliferation of multiple Farcaster clients is one of our motivations for /concept-testing Instead of founders trying to compete against Warpcast on their own, we can collaborate on open-source projects to help new clients flourish https://warpcast.com/jrf/0x2b238b6c

Now, to the details!

Level 1 Cozycast

$500-$2000 USD in crowdfunding raised

The basic requirements include:

1. A web & mobile app (Android + iOS)

2. A default "Following" feed

   • All casts from people you follow, in the channels you follow

3. The ability to create custom feeds with additional variables:

   • Casts restricted to users with certain FIDs

   • Casts restricted to users with a specific NFT in a connected wallet

4. The ability to open casts in Warpcast

   • Cozycast is "read-only" by design, so in order to cast or interact with frames, it will need to be part of an app with signers.

Level 2 Cozycast

more than $2000 USD in crowdfunding

5. Additional filters for custom feeds:

   • Show casts only in specific channels

   • Show all casts vs. just people you follow

If we raise significantly more than $2000, we may consider additional "bonus" filters such as muting certain keywords.

Rationale for this MVP

The goal with this scope is to keep things practical.

• What are the most important features to make the concept viable?

• How much additional work will each feature or requirement add?

• How large of a bounty would be required to get this work completed?

We are trying to validate the concept of Cozycast, not build a fully functional product.

Next Steps

The crowdfund will be live for one week – closing on March 27th at 10:50am ET. (Next time I'll launch at a rounder number.)

Afterwards, Samuel will launch a bounty with the detailed requirements and some guidance for interested builders. The first party to complete the bounty will win 90% of the raised funds. (We keep 10% to fund Concept Testing.) If nobody completes the bounty within a month all of the raised funds will be returned.

This is our first experiment in a sufficiently decentralized process. We know there are other ways of funding and building (such as doing them in the other order!) but we are committed to building a scalable way to contribute to the Farcaster ecosystem.

If you're interested in participating in future projects, feel free to reach out to Samuel or myself or post in the /concept-testing channel.

As a reminder of how all this works, you can read our announcement post.

Alright, on to the concept!

What is Cozycast?

Cozycast is a way to filter the firehose of Farcaster into feeds that matter to you.

We envision a simple app that is designed to let users stay focused on the people, channels, and communities of their choice – free from additional noise.

Using a mix of the Farcaster social graph and onchain data, Cozycast users will be able to create feeds to their own liking.

Why Now?

The recent explosion in user growth has transformed the experience on Farcaster. For a long time, you could pretty easily see whatever was posted by the people you followed. Starting at the end of January, the number of daily users spiked, reaching over 40,000 unique casters at the peak. To put that growth in context, the network only registered 2,000 daily casters (95% fewer people!) once in December.

The Warpcast team did an admirable job moving quickly to prevent spam and “low quality” casts from appearing on the main feed. However, the amount of content on Farcaster grew by an order of magnitude in a relatively short time...

Farcaster

six — q/dau 🎩 on Warpcast

this place has changed forever

In the early weeks of this new era, overwhelmed users retreated to private group chats (only released in mid-January themselves!) and ventured out onto the frontier with niche channels.

Both of these behaviors will continue, but they won’t solve the core challenge of curating the conversations on Farcaster.

Warpcast continues to work on moderation tools, but mainly for channel hosts. For the individual user, there aren't any great tools to curate the stream of casts. As of now, there is no way to mute specific channels – you will see posts from people you follow even in channels you're not a part of. (This is great for discovery, but not for coziness.)

Farcaster is only going to keep growing – which means more users, more communities, and more noise to sift through. And crucially, noise is different for everyone.

What is the Solution?

Fortunately, with a sufficiently decentralized network, we can build new products that are designed for a different browsing experience, using the same underlying data.

For example, Dan has mentioned repeatedly that someone could build an app that only shows casts from the first N accounts.

However, FIDs are just one variable that can be used for curation. Other onchain data, such as token balances, NFT ownership, and specific network activity, may be useful for filtering feeds for different use cases and preferences. Farcaster data, such as follows, casts, and channels, aren't technically onchain but are available through Hubs.

It's easy to imagine a feed for /purple only featuring casts from those who actually hold a Purple NFT. Or a feed that focuses on prolific Zora creators. If you're a creator of onchain content, you could create a feed that features people who have collected your work. In the future, custom feeds could leverage some sort of decentralized reputation score.

Cozycast lets people design feeds however they want.

Scoping an MVP

Concept Testing is designed to get ideas from 0 to 1. We want to test concepts based on the core idea instead of launching fully-featured products.

For Cozycast, that means a relatively minimal experience focused on custom feeds. It is designed to be used as a complement to robust clients like Warpcast.

An MVP would likely be read-only (link out to Warpcast) and have no notifications. (Does this make it cozier?!)

Reminder: since Concept Testing projects are open-source, this wouldn't necessarily be the end state. If a concept gets funded and built, anyone can keep working on it or incorporate the code into their own project!

The specific details will be hammered out in a working group before being finalized for a crowdfund and bounty.

If you'd like to be part of that group, collect this post!

Parting Thoughts

This is not the first time that someone has explored the idea of custom feeds. In fact, David Furlong (@df) built Discove more than a year ago to solve a similar problem.

David discussed some of the reasons for pivoting / shutting down the project here.

The major difference between Discove and Cozycast, in my eyes, isn't the product but "business model". Discove was a founder-led project that ultimately required a steady source of revenue to continue development. Cozycast has no "founder", will raise funds before development, and be open-source for anyone to continue working on or plug into their own project.

We're looking forward to working with founders on these projects. Instead of incubating competitors to nascent products, we believe that Concept Testing can be an asset to builders with similar ambitions. Most obviously, we can fund development that founders can freely implement into their own projects. But we can also drive awareness of a problem and highlight those that are working toward on solutions. Open-sourcing our work makes the Farcaster ecosystem resilient: it "raises the floor" for users by ensuring that these necessary functions will always be available, even if founders decide to pivot or sunset their products. However, the best experiences will always be built by dedicated founders that obsess over their products and listen to users.