As a way to organize our thoughts around moderation, it may be helpful to categorize users into a few commonly seen categories. The following are five types of users I routinely recognize:

Blatant spammer. These accounts just spew random stuff into your channel.

Engagement farmer. These accounts pretend to post on-topic, quality content while actually just providing low-effort stuff intended to maximize engagement.

Quality contributor. These are the accounts that get everything just right.

Oversharer. These accounts could be quality contributors in principle, except they post too much.

Newcomer/guest. These accounts may be contributing interesting material but they are too new or cast too infrequently for the moderators to know and recognize them.

The main job of moderation is to elevate the quality contributors and the interesting contributions of newcomers or guests while rate-limiting the oversharers and not being tricked into elevating the engagement farmers. In a moderation system focused on banning/hiding, you would try to identify the blatant spammers and engagement farmers, ban those, hide some portion of the posts from the oversharers, and let everybody else post at will. Farcaster is doing the opposite currently and emphasizing curation rather than hiding. In this context, you will want to elevate the casts of quality contributors, as well as the interesting contributions from newcomers and some fraction of the material from oversharers.

The blatant spammers are easy to identify. Their casts are obviously inappropriate or off-topic, and their cast volume tends to be high. The engagement farmers on the other hand are more subtle. On first glance their casts may seem reasonable. I've found this to be particularly problematic in /ai-art, where people just spin up ChatGPT to generate a constant stream of motivational quotes with accompanying images. The resulting casts can look very convincing but once you've seen a couple hundred of them you recognize them right away.

The quality contributors are typically the reason you visit a given channel. Their casts tend to be on topic and interesting, and they cast somewhat regularly yet not so much that they dominate the channel. The oversharers don't look that different from quality contributors except they post too much. If you don't slow them down, they'll just take over the channel and drown out all other material. Finally, the newcomers or guests tend to not be known to the moderators, so their casts need to be evaluated carefully on a one-off basis.

What are best practices to implement moderation? In principle, a single person could moderate a channel, but in practice I think it's too big a job for one person, even for a small channel. You need a team. And even then, people get distracted and have other things on their plate and forget to moderate so you may want to automate some moderation. One thing you could consider is allowlisting. Maybe all moderators could be allowlisted. If you can entrust them with moderating Main you can probably also entrust them with casting into Main. Next, you could consider identifying specific quality contributors and allowlisting them. There's a tradeoff here between the size of the allowlist and the amount and type of work the moderators have to do. The smaller the allowlist, the more moderators have to find individual casts in Recent and curate them, and the larger the allowlist, the more moderators have to monitor Main and potentially hide casts from there. Which approach is more appropriate may depend on the channel size, the moderation team, the composition of the channel audience (small, close-nit group of frequent contributors versus large community of occasional contributors) and also simply personal preference of the channel owner.

I also think allowlisting could be used more aggressively if we could apply a rate limit to allowlisted accounts. For example, each account might be allowed to cast two casts per 24 hours into Main but any further casts in that period would be hidden by default (but could be rescued manually by moderators). This would be a great way of moderating oversharers, as they could be allowlisted and at the same time rate limited. This is not currently possible but the Farcaster moderation tools are under heavy development and I hope this type of moderation will become available eventually.

I'd like to point out that the default Warpcast moderation setup, where casts from accounts with power badge are always shown in Main, approximately implements the strategy I have outlined here, by making the assumption that all holders of a power badge are quality contributors and should be allowlisted. And I'm afraid this will frequently not be a good assumption. Even if somebody is a quality contributor in aggregate, taken over their entire activity on Farcaster, they may be an oversharer, engagement farmer, or simply a confused guest in one specific channel. So I feel confident allowlists will generally have to be constructed specifically for individual channels, or maybe for families of channels. (Various different art channels could share allowlists, for example.) We definitely need better tools to construct such allowlists if we don't want to do this entirely by hand.

Let me end with some thoughts about moderating /ai-art specifically. Since it is so easy to generate AI images, this type of a channel can get overrun with tons of low-quality or low-effort material. Now of course the moment I say this, people will complain that effort is irrelevant for art. Somebody can write a three-word prompt that produces outstanding art while others may put tons of effort into their work but the outcome is mediocre or irrelevant. I agree. That's why I tend to evaluate effort at the level of the account rather than the individual cast. I try to determine whether the individual has a consistent artistic practice or instead is just engagement farming on Farcaster. I won't specify in detail how exactly I do this, but suffice to say I look for consistency, a presence outside of Farcaster, engagement with other artists I know, discussion of methods, techniques, or concepts expressed in the art, and so on. Accounts that lack all of this and instead just constantly post their latest AI-image mint on Zora are unlikely to score high on my "consistent artistic practice" scale.

Of course any one assessment I make may be wrong or biased. That's why we have a moderation team. I may have a bias in favor or against certain accounts and others on the team may have the opposing view. That's great. I don't think moderation has to be particularly strict to be effective. In case of doubt, it's better to let more content and diversity into Main. There's still the added level of moderation coming from Farcaster's algorithmic sorting. It gives all the channel subscribers the power to increase or decrease the visibility of certain casts, by liking and interacting with them on the one hand or ignoring them on the other. So, as you're browsing the feed, please don't think that because there is now manual curation you no longer have to interact with casts. Please continue to like and recast the best casts you come across. It will help their authors and it will improve the overall feed quality on the platform.

Decentralized channels come with decentralized moderation tools, and so there are multiple options you can employ. The two most prominent ones currently are automod and airstack. I personally have only used automod so that's the system I'll describe here, but I would expect airstack to work similarly, at least in terms of the available moderation primitives. It is relatively straightforward to switch out your moderation engine, so it's fine to get started with one and at a future point switch to another if it provides additional features or benefits or lower costs. (The tools are free at present but will eventually require payment as they have ongoing maintenance costs.)

All of these tools fundamentally work in the same way. They employ a bot that is given moderation privileges for your channel. You interact with the bot to moderate. The interaction with the bot can be purely manual (the bot won't do anything unless you tell it to), or purely automated (the bot implements a set of moderation rules you have defined), or a mix of the two. When you hear "bot" you may initially think of purely automated moderation and you may not like the concept. I think it's best not to think of this type of moderation early on. Just start out in purely manual mode and then add automated rules as the need for them becomes apparent.

Why do you need a bot in the first place? You don't if you as the channel owner want to do 100% of all moderation yourself. I assume you don't. (You may want to sleep sometimes.) The new decentralized channel setup allows for one moderator account which can be distinct from the channel owner. So, you can do moderation yourself, delegate it to a trusted friend, or delegate it to a bot that provides a multitude of features, such as moderation teams, token gating, automated rules, etc. For all but the smallest channels people will likely opt for the bot setup.

There is one last concept you need to know before you can think about creating your own setup. What does the bot actually do when it moderates? I.e., how does moderation work in the new setup? Moderation is a simple binary flag, indicating whether a cast should be included in the Main feed or not. And this binary flag is a like of the cast by the moderator account. In other words, if the moderator bot likes a cast the cast will be shown in Main and if the bot doesn't like the cast (or removes a like it has previously applied) the cast will be removed from Main. That's it. Actions such as curate, ban, hide, etc. all just manage how the bot likes the casts in the channel. For example, "curate" means the bot applies a like and hence sends the cast into Main. "Hide" means the bot removes a like (if previously applied). "Ban" means the bot no longer applies likes to casts from the banned account, potentially overriding other rules that would imply casts should be liked. And so on.

So let's talk about turning these general concepts into a concrete automod setup. Why automod? I had a quick look at airstack, and it seems a much more powerful and complex tool, with many different things it can do beyond channel moderation. By contrast, automod has a singular focus on moderation. Therefore, to start out, I assume you'll be better off with automod. (I have never used airstack though and would be interested in reading a blog post on how to set it up from start to finish.)

Disclaimer: I am not a member of the automod team and anything I write here is purely based on my experience and experimentation with the system. Some of the things I say may be wrong. Some things may have changed between the time I wrote this and the time you read this. This post is accurate to the best of my knowledge as of June 2, 2024.

Let me also say that automod is extremely minimalistic, to the point where in several instances there is literally a single button you can press, and you either do so or you don't. It works, but it doesn't spend many words explaining how things function and what exactly different parts of the setup do. If anybody from the automod team is reading this, you guys could benefit from a little more text that explains to your users what is happening.

To get started with automod, you go to automod.sh and log in with Farcaster. Then, you get the option to set up a new bot. You need one bot per channel you are moderating. For now, just click on the button to make a new bot, enter the name of your channel, and leave all other settings as is. The default setup mimics channels without moderation, where casts from users with a Power Badge make it into Main and all others do not. After you have created the bot, the automod site will remind you that you have to set the bot as the moderator account for your channel on Farcaster. If you don't do this, or don't use the correct bot name, nothing will happen in terms of moderation. Make sure you copy the correct bot name. (The automod system runs multiple bots with different names for efficiency.)

Next we can configure the bot we have just created. Let's start with "bypass", which I think is one of the most important settings to keep things manageable. (You can find bypass underneath the filtering rules in the Rules tab.) Through the bypass mechanism, you can set up a list of accounts whose casts will always make it into Main. This allows you to set up a list of trusted channel contributors that don't require manual approval each time. Because it's easy to expand or shrink this list, I think it's a very powerful tool for manual curation.

For my AI-art channel, all we use at present is bypass and manual curation. However, I have found that bypass doesn't work unless I set up some rule that simply hides all casts in the channel. I created this rule by asking automod to look for a rather long and arbitrary string of text in the user profile and hide all casts that don't contain this string of text.

I will say that automod is a bit confusing as of this writing because the available actions are negative (hide, ban, cooldown) rather than positive (curate into Main). But, because there is an "invert" button that converts each rule into its opposite, you can pretty much express any rule you want with the tools provided. Also, my understanding is the negative actions are a holdover from how moderation worked a few weeks ago and may soon be switched.

While we're not using automated rules in AI-art, I will point out that automod has a wide range of possible rules, so chances are whatever type of automated criteria you may want to set up (token gating, only allowing certain users to cast, etc.) automod can likely do it. It is also possible to combine multiple rules for even more complex moderation scenarios.

Now let's move on to setting up manual moderation. Automod allows you to create multiple groups of moderators with different abilities to moderate. Maybe some can only curate into Main, while others can curate and hide and ban. At present, in AI-art, I have a two-tiered setup where one group of moderators has access to everything (including the automod settings, see below) and a second group has more limited ability to moderate (currently they can curate, hide, and issue a cooldown).

It's easier to grant privileges than to take them away, so my recommendation would be to start out slowly and initially give new moderators only limited access to the moderation system and once they prove themselves as judicious moderators you can give them increasingly more responsibility.

Along the same lines, I think it's a good idea to have a small team of trusted collaborators that help you manage a channel and its automod configuration. Automod allows you to name such collaborators, and they will have full access to all automod settings.

Finally, I would like to mention that automod provides a detailed log of all moderation actions that have been taken, so whenever a cast either makes it into Main when you think it shouldn't or didn't make it into Main when you think it should, you can check what exactly happened. This also means that should a moderator on your team go nuts and inappropriately curate or hide all sorts of casts you would be able to see this and revoke their privileges.

This completes the automod setup, but we still need to talk about how exactly manual curation is effected. It works via cast actions, which means all moderators need to install the required actions in their Warpcast setup. Automod provides convenient install links for those actions on its moderators settings page, but most of your moderators won't have access to the automod setup so this won't be of use to them. So, you will have to tell them to go to the Warpcast cast actions site, search for "automod", and then install "Curate", "Hide", and whatever other actions you have enabled for them. Then, to curate a cast, the moderators click on the actions button of the cast in their feed and then select the relevant action. This will typically require two clicks, so is rather quick and nonintrusive. You can easily moderate a bunch of casts while you're reading the feed.

I have just been made moderator of a channel. What now? How do I moderate? The first thing you have to do is install the relevant cast actions. Most importantly, you probably need "Curate", which you can find here. Additional cast actions are available here (just search for "automod"). Depending on how you have been set up, you may have access to "Hide" "Ban" "Cooldown" etc.

If one moderator has curated a cast, what happens if another curates the same cast? The second curate has no effect. The curate action simply causes the automod bot to like the cast and the bot will not like the cast a second time. If you want the cast to be more visible like it from your personal account.

Is there an easy way to see if a cast has been curated already? Yes and no. Yes in the sense that you could look up the likes of the cast and see if automod has liked it and if yes then you know the cast has been curated. No in the sense that this procedure requires multiple clicks and is much more cumbersome than just curating the cast again just in case. That's what I do. I just curate what I think should be curated and I don't worry whether a cast was already curated by another mod.

Will the author of a cast know their cast has been curated? They will see that their cast was liked by automod. They will not know why this happened. (For example, in a channel with multiple moderators, they will not know who triggered the curation.)

What happens if one mod curates a cast and another mod bans the same cast? The last action will take effect. "Curate" means automod likes the cast. "Ban" means automod removes any like it may have previously assigned. You can keep applying these actions to the same cast as many times as you like, and whatever is the most recent action is what will determine whether the cast is visible in the Main feed or not.

Is there a public list of all accounts that can moderate a channel? Not at present, as far as I can tell. The channel owner would have to separately assemble the list and publish somewhere.

Can moderators affect the ordering of casts in Main? No, the ordering is currently determined by the Warpcast algorithm. However, by liking or otherwise interacting with a cast you increase its visibility and the likelihood the algorithm will move it higher up in the feed. Yes this is somewhat dissatisfying. It's what we currently have.

Can moderators see the raw unfiltered feed for a channel? Apparently this is currently not possible except for the channel owner. This is one of my biggest gripes with the current system. Moderators need to be able to see the raw feed to discover gems from newer or unknown accounts or accounts not in their following list.

The new system represents an approach to moderation that has never been tried at scale on a social media site, and I believe it has the potential to become a defining feature of Farcaster. To explain this argument, let me quickly discuss the alternatives to channels that are currently available on other sites. The closest equivalents are probably on one extreme hashtags on sites such as Twitter/X, Instagram, or TikTok, and on the other subreddits on Reddit. Farcaster channels live somewhere in between, but also are very different.

Let's first consider hashtags. The defining feature of hashtags is that anybody can add any hashtag to their posts and thus insert their content into the hashtag's feed. This encourages spamming. The raw feed of any popular hashtag would be completely unusable. So social media sites engage in extensive filtering and algorithmic curation. They tend to elevate the most engaging content and suppress anything that appears spammy, inappropriate, or otherwise bad for their business metrics. Importantly, there is no human curation. Hashtags don't have individual moderators that can tweak the feed or the filtering criteria. And there is typically no filtering for on-topic posts. If engaging but otherwise off-topic material starts getting pushed into a particular hashtag the hashtag quickly becomes useless. More generally, what a social media site considers to be engaging or spam or good/bad for their metrics may not be aligned with what the site's users would actually like to see under the given hashtag. Consequently, to the site's user, hashtags often don't deliver the value they promise.

Now let's consider Reddit. It attempts to combat the problems of spam and off-topic posts via a combination of up/down voting, human moderators, and automated moderation tools. If you have posted to Reddit, you know the first hurdle is to make it past the automated filters. Your post may be too long or too short or contain the wrong keywords etc. and get rejected outright. And once you're past this filtering stage, your post may yet get deleted by a human moderator because it's a duplicate or off-topic or just not what the moderator wants to see. And once you're past this gauntlet, your post needs to receive many more upvotes than downvotes to be visible in the main feed of the subreddit you're posting to.

In theory, the extensive amount of filtering that Reddit applies should result in a very high-quality feed. But in practice there are obvious problems. First is the hivemind combined with the ability to brigade the voting. It happens frequently on Reddit that constructive and perfectly valid posts or comments get downvoted because they don't align with the preferred perspective of the subreddit or because there is a coordinated attempt to suppress this particular viewpoint. Second is the emphasis on human moderation via removal of undesirable content. Reddit mods have massive power to remove content but almost none to bring content to the fore. So their entire mindset is being shaped into acting like the thought police. They spend the majority of their time engaging with the less desirable content and thinking about what should be removed rather than what should be amplified. This dynamic, I believe, selects for a certain type of person who likes to tell other people what not to do. If you enjoy finding interesting stuff but don't enjoy policing people's behavior you're not going to have fun as a Reddit mod.

So how is moderation going to work in Farcaster channels? First, we need to acknowledge that many strategies employed by centralized platforms don't work in a decentralized system. Most importantly among those are banning and hiding. In a decentralized system, anybody can write any content into a channel. You could ban users or hide casts in a particular client, such as Warpcast, but this would have no effect in other clients. In fact, I frequently use Supercast to find casts hidden by Warpcast. The more people use alternative clients, the less what Warpcast does in terms of moderation matters, so we need a different approach if we want channels and channel moderation to be a feature of the entire Farcaster ecosystem and not just a Warpcast-specific feature.

The strategy that does work, however, is to specifically tag the casts that are the most interesting and display those in a separate feed. That's what the main feed does in Warpcast, and other clients can emulate this strategy and provide a very similar curated feed to their users. In this setup, the job of the moderators is not to flag the bad content but instead to elevate the good content, i.e. to curate. Of all the casts sent into a channel, which are the ones that deserve the most attention, that subscribers to the channel should see first? It will be the moderators' job to find those casts. Channel hosts and moderating teams can come up with a variety of strategies to do this curation task, and different strategies may work for different channels or topic areas or teams. We will see over time what the best practices are. How exactly this works on a technical level is beyond the scope of this article, but rest assured that the system is very flexible and powerful and chances are whatever moderation strategy you would like to employ will be possible.

The main reason why I am so excited about the new approach is that I believe manual curation is an important strategy that has been lacking in social media. Curators can develop a specific vision of what kind of experience they want to create and then distill that vision in their moderation approach and channel. This will ensure that each channel will have its own unique culture and content mix. Sure, some will be terrible, but others will be really interesting and way better than any automated system would deliver. Voting systems (like on Reddit) and algorithmic curation (like on X/Instagram/Threads/...) tend to lead to regression to the mean and result in relatively uniform, boring feeds. Manual curation has the potential to create the opposite.

Some people will complain that the new system puts too much power into the hands of the curators, who may be gatekeeping or overly opinionated or simply ruining a channel with their personal curation choices. But there is an important counteracting force: Anybody can create a channel. If you don't like how a channel on a given topic is curated you can set up a competing channel and moderate differently. I believe opinionated is good, because high quality content is always opinionated. And, if a channel gets moderation entirely wrong it'll lose out over time to another channel that gets it right. The channels that consistently provide the most interesting main feed will grow the fastest.

If you want to learn more about the technical aspects of how the new moderation system works and is implemented, I encourage you to watch this GM Farcaster episode from May 28, 2024:

I also plan to write a follow-up post that will explain how the new system works from the perspective of a channel host and how to set up and use the new moderation tools, and specifically automod.

This article was previously published on fx(hash).

Generative AI for image creation has divided the world into two camps. On the one hand, there are people who are enthusiastic about this novel set of tools, who incorporate AI into their artistic practice, and/or who collect works from the people who do so. On the other hand, there are people who consider AI-generated imagery low effort, not art, and/or an affront to visual artists who have spent years honing their craft. I am squarely in the camp of the AI enthusiasts, and here I would like to argue in favor of this position.

One common critique is that AI is inherently predatory, stealing works from other artists. Let's get this critique out of the way first, as it is not actually inherent to AI and this article is not about the topic of potential copyright violations in AI art. I think we can all agree that if the AI was exclusively trained on images that were explicitly licensed for such use then there is no issue. The training images could be all in the public domain, or a large image repository such as Getty Images could decide to train an AI on their image catalog, or an artist could train their own AI exclusively on images they have taken themselves. (Ivona Tau does exactly this.) So for the remainder of this article, let's assume the artist has both the legal and the moral right to use the AI to generate images for commercial use.

The more interesting question, from my perspective, is whether AI-generated imagery can be art. To address this question, we need a definition for what is art. I'll give you mine:

A person is making art whenever they put more care, craft, skill, or thought into something than is required from a purely utilitarian perspective.

By my definition, a sushi chef who has honed their skill of cutting rolls to exact dimensions is creating art. A scientist who carefully refines and polishes their visualizations or their writing is creating art. A gardener who carefully plants, trims, and prunes may be creating art. You may not agree with my definition and find it overly broad, but if so I'd invite you to consider why you want to deny so many people their artistry. Do you lose anything if a chef, a scientist, or a gardener is an artist?

If you allow me to proceed with my definition of art, it directly leads to the question of how much care, craft, skill, or thought AI practitioners put into their work. And in fact, the most commonly raised critique against AI imagery is that it is low effort. Now it is certainly possible with AI to create images without putting in much effort, if you don't particularly care what you get out. Provide a few quick prompts to DALL-E or Midjourney and they'll hand you some images. But just because a medium can be used this way doesn't mean it's the only way it can be used. I can make low-effort collages, but nobody would doubt that there are true collage artists that create amazing works. I've never heard anybody accuse Wangechi Mutu of producing low-effort art, and yet many of her works are just cuts from magazines glued onto a canvas.

One way for an artist to get around the low effort critique is to build and train their own models. Many of the AI OGs do that, because as little as a few years ago this was the only way to create images with AI. Let's consider Ivona Tau once again. She takes a collection of her own photographs, uses them to train a neural network that she has coded herself, and then uses the trained neural network to generate unique artworks that reflect but also transform the original photographs. Whatever you think of her work, it is not low effort, that's for sure.

However, in my opinion, while training your own models is laudable, it is not a requirement for making meaningful AI art. Let's consider what people do who use primarily or exclusively some of the widely-available, off-the-shelf models such as Midjourney, DALL-E, or StableDiffusion. First, they may spend extraordinary amounts of time giving the model different prompts, over and over, to explore the model's latent space, that is, to determine exactly which prompts have what effects. In addition, they may explore how to fine tune prompt weights to get specific outputs or compositions, how to modify model parameters, which sampling algorithm to use, and so on. It is not unusual to generate hundreds or thousands of outputs to fine-tune everything just right for one final image. Second, in addition to basic prompting there are so many other tools these artists can use to gain more control over their creations. There is image-to-image, upscaling, inpainting and outpainting, ControlNets for spatial conditioning, the list goes on and on and more tools are added every day. The AI space provides plenty of room to make this a high effort activity, and in fact I would like to invite you to try this out for yourself to see how much effort and skill is actually required to get images you are satisfied with.

Another critique I have seen is that the artist has no control over the image output and composition and is just randomly sampling possible images from the total set of outputs the AI can produce. Well that's certainly not true at all for image-to-image approaches, where the artist provides an input image that guides the AI. This is a commonly used technique. My collection AI Artifacts was made this way (see below). If you compare the AI output to my input image, you can see that the AI followed my image prompt quite closely and mostly provided texture and depth. If I had created texture and depth by gluing magazine clippings onto a canvas, Wangechi Mutu style, nobody would bat an eye and accept my work as art. But if I instead use a generative AI model that doesn't count? The AI was somehow able to take my code-generated artwork and turn it into non-art? That's quite a feat.

Finally, I've seen some critics draw a sharp distinction between the people that write their own neural network code and/or train their own models and the people who use off-the-shelf models, arguing that the former may deserve the label of artist but the latter do not. Of course everybody is free to draw their own boundaries but I cannot at all agree with this assessment. From my observation, people using off-the-shelf models produce works at widely differing levels of quality. It's not all the same "low effort" stuff. Instead, some people consistently amaze me with what they can coax the AI to do, while others produce work that is much less impressive. Seeing these differences I feel comfortable concluding that there is true artistry and skill involved at the upper end of the performance scale. Not everybody is equally talented or capable when it comes to working with AI.

A great place to observe this difference in skill levels is the long-form AI platform EmProps. Everybody who publishes on this platform has access to the exact same handful of models and there is neither manual post-processing nor (for the long-form drops) manual selection of outputs. The EmProps OpenMarket Beta has released several hundred long-form collections. I'd like to encourage you to explore the different artists and see to what extent they get very different results using the same tools and methods.

In my mind, working with generic AI image-generation models is almost the purest form of artistic expression. Everybody has access to the exact same tools. The ability to generate imagery is not limited by somebody's physical ability to move a pen or brush in a precise manner or by their mental ability to write sophisticated graphics rendering code. Instead, all that matters is artistic intent, imagination, and judgment. In the musical arts, we are perfectly happy to celebrate composers who have written masterpieces, even if they never play any instrument themselves. I believe we should similarly celebrate the most impressive AI artists. They are visual composers who tell stories, create emotional reactions, or otherwise delight us with their artistic perspective and vision, even if they never actually draw, edit, or touch up the images they produce.

If you enjoyed this article, you may also like my earlier article discussing how little AI is necessary to call an artwork AI art.

I made an end-run around an AI generative art platform to release art that doesn't actually use AI for its image generation. However, because I couldn't entirely turn off the AI, it leaves a trace in visible image artifacts. Is the result AI art? Is it not? Should you collect it if you collect AI art? Should you collect it if you don't collect AI art but do collect code-based art? Read further to learn everything about this collection.

The long-form generative AI art platform EmProps recently opened its OpenMarket Beta, allowing a larger number of artists to explore the platform, mint their projects, and see what they can do within the framework of long-form generative AI. In many ways, the OpenMarket Beta feels like early days of fx(hash), with tons of releases, priced extremely cheap, many collections not minting out or minting slowly, and plenty of platform bugs that lead to slow reveals, occasionally missing tokens, and everything else you can imagine happens when a new platform first starts scaling.

I got my hands on a beta membership and started exploring also. And while I was exploring and talking to people, I was asked whether EmProps mandates that all released art must be AI. Thinking about this for a bit, I realized that even though the EmProps minting pipeline tightly integrates the AI model Stable Diffusion and we can't turn it off, we can actually choose Stable Diffusion settings that mostly (but not entirely!) make an end-run around the AI. And then, in combination with EmProps' ability to generate code-based images using the p5.js framework, we can make traditional code-based long-form generative art on EmProps. This line of thought let me to create and release the collection "Is it AI art if a human designed it?"

Here, I'll explain in detail the ideas behind this collection. To start, let's review the basics of generative AI art. There are two main approaches, text-to-image and image-to-image. In text-to-image, we provide the AI with a written description of the type of output that we want, and the AI generates an image based on this description. In image-to-image, we provide a written description and in addition we also provide an image that serves as a visual guide for the output. You can see in the example I have provided how in the text-to-image approach the AI freely chooses the colors and geometric arrangement, but when we use the same prompt in image-to-image the output is now very different and clearly constrained by the provided image.

Importantly, the image generation process is in part driven by random noise, which means the AI can generate many different outputs from the same input prompt. The EmProps platform enables artists to create long-form collections of AI art by creating many different images from the same prompts on the fly as the user mints the art. (EmProps also allows artists to use different text or image prompts within the same collection, for added variety, but this is less relevant for the present topic.)

In the image-to-image mode, there doesn't need to be any particular relationship between the type of prompt image used and the desired output. For example, above I used a photograph as prompt to generate abstract oil paintings. Similarly, I could use a hand-drawn sketch to generate photorealistic images. Because the AI can produce interesting outputs on the basis of virtually any prompt image, a popular technique is to use code-based generative art as prompt. The code-based prompt images don't have to be particularly complex or sophisticated, as the AI will handle textures and subtle color variations. This technique is particularly attractive for long-form AI art, where we can use a different, code-based image prompt for each output. I call this "generative image-to-image," and the EmProps platform supports this mode also. We can upload any sketch written in the p5.js framework and EmProps will run the code with a different random seed for each iteration, capture the generated image, and use it as image prompt for the AI.

Whenever we use image-to-image, we need to set a parameter called "denoising strength" that determines how closely the AI generated image adheres to the image prompt. At a denoising strength of zero the output image looks nearly (but not exactly) identical to the prompt image. At a denoising strength of one, on the other hand, the AI has wide latitude to deviate from the prompt image. Colors, arrangements, textures, types of objects shown, anything can change with a denoising strength of one. The term "denoising strenth" may sound weird but it makes a lot of sense when you look into what diffusion models such as Stable Diffusion (and Midjourney, DALL-E, etc.) do. They generate novel images by attempting to successively remove noise from initially random input. So, a denoising strength of one means maximum amount of noise removal relative to the input image, which means maximum creativity. By contrast, a denoising strength of zero means no noise gets removed, and we get the input image back as our output.

You may wonder why I said that at a denoising strength of zero the output image is not exactly identical to the prompt image. In my example provided here, they certainly look the same! To understand this, we need to know a little more about how diffusion models operate. They all have a compressed internal image representation called the "latent space." This latent space does not store every single color of every single pixel of the output image, but instead stores something we can think of as a description of what the image should look like. You can imagine it providing descriptions such as "there are sharp blue-yellow edges in the bottom right of the image, and a green-blue gradient across the top, and something circular and red in the middle." The AI then uses a system called a decoder to turn this latent space description into the actual output image.

When Stable Diffusion generates images, it operates entirely in the latent space. Because the latent space is much smaller than a typical image (think about 50 times smaller), computations in latent space are more efficient. In fact, the idea of operating in latent space rather than on images directly was the main novel contribution of the original Stable Diffusion model. So how do we get our image prompts into the latent space? For that, Stable Diffusion uses a module called "encoder." It is the inverse of the decoder. The encoder takes an image and converts it into latent space, and the decoder takes the latent space representation and converts it back into an image. The combined system of encoder, latent space, and decoder is also called a "variational autoencoder" in the machine learning literature. You can learn more about this latent space and how Stable Diffusion works in detail from this article.

If you've followed so far, you should ask yourself whether any arbitrary image can be accurately represented in latent space. If the latent space is 50x smaller than the actual image, surely there must be something that gets lost in the encoding? This is indeed the case. While latent spaces are surprisingly good at encoding images, some fine details or subtle aspects of the image may get altered or distorted. This is very similar to what happens in lossy image compression methods such as JPEG. JPEG is great for storing images, but when you turn the quality slider all the way down to get a small file size you can end up with ugly artifacts. For more detail, see this Wikipedia article or the chapter on compression of bitmap graphics in my dataviz book.

So now we can understand what happens if we use the image-to-image generation method and set the denoising strength to zero. The image prompt gets encoded into the latent space, Stable Diffusion retains this exact latent space representation, and then the decoder converts it back into the output image. So in practice we haven't any done diffusion or image generation. We have just compressed the image using the AI's latent space.

When we combine this latent encoding step with traditional code-based long-form generative outputs, the result is code-based art that uses AI but the AI has no meaningful influence on the design of the final outputs. It just adds latent space artifacts. Conceptually this is similar to taking the code-based outputs and storing them as JPEGs with low quality settings. This is exactly what my art project "Is it AI art if a human designed it?" does. All the key visual features of the outputs, such as colors, arrangement, background textures, blurriness, etc., are decided by the p5.js code. The AI only adds minor artifacts. If you look closely, you can see that the AI contribution consists mostly in distorting small letter shapes and making some lines sharper than they were in the p5.js output. The following animation shows the before/after of one whole image and of various parts of the image zoomed in.

Unfortunately, on EmProps we cannot see the p5.js output for individual iterations, we can only see the final outputs. So you have to somewhat take my word for it that the p5 outputs look just the same. To make this argument plausible, here I have provided a set of ten outputs straight from p5.js. At first glance, the outputs look just like what you see on the EmProps platform for the final mints. However, if you zoom in, you can see that all the letters are properly shaped and look identical. There are no AI artifacts in these images.

There is one additional step in the AI pipeline that I haven't mentioned yet. Diffusion models typically generate relatively low-resolution images, on the order of 1000x1000 pixels. So to obtain images with the resolution we would expect for modern displays or for printing, we take the generated images and then upscale them, using yet another AI model. How exactly upscaler models work is beyond the scope of this article. Suffice to say that they aim to reconstruct what a higher-resolution version of the same image would have looked like, and this works well if the image looks similar to images the upscaler was trained on. For my particular artwork, I find that most artifacts are introduced at the Stable Diffusion stage, and the effect of the upscaler is mostly limited to making the overall image sharper.

Now you know everything you need to know to understand the collection. What you do with it is up to you. Is this AI art? Is it not? Should people interested in AI art care? Should people interested in code-based art care? Should anybody care? I'll leave this to you. As of this writing the collection is only about half-minted, so if you like it you can still mint your very own iteration right now, over here.

What prompted me to write this post is I just listened to the GM Farcaster interview with Jacek and attention farming didn't come up once, even though there were extensive conversations about other types of farming, about spam, about increasing the quality of tipping, and so on. So maybe people are not fully aware of the dynamic that has been playing out on Farcaster for the last couple of weeks. Therefore, let me describe what I mean by attention farming, why I think it's bad, and what can be done to reduce the incentive to engage in it.

I define attention farming as the behavior of deliberately using your $DEGEN tip allocation to increase interactions with your casts with the goal of increasing your $DEGEN tip allocation. This is different from regular $DEGEN farming as the primary goal here is not to receive more $DEGEN tips (though this may happen as a side-effect) but rather to receive more attention and a larger tip allocation. Almost everybody who has or has had a meaningful $DEGEN allocation has engaged in it, including myself.

These farming casts generally have the form of asking for a specific behavior in return for $DEGEN. A common example is a contest: "Reply with your best meme about coffee. 5000 $DEGEN for the tree top responses." Another common example is a cast that rewards all responses: "Reply with your current works in progress. 200 $DEGEN for all on-topic responses." It seems obvious to me that many people have figured out that these casts generate so much engagement that they meaningfully raise one's ranking and thus $DEGEN allocations, and therefore spending the $DEGEN allocation on this type of cast is the most rational thing anybody can do. Moreover, and this is particularly insidious, for somebody with a large daily allocation it can be difficult to give away the entire allocation every day. Therefore, farming casts may seem like the most fair or most constructive way of dealing with this situation.

Now you may wonder how attention farming differs from regular marketing. What's the difference if I run a contest and reward the best responses with, for example, USDC or ETH? I see at least two differences. First, marketing expenses come with a real opportunity cost. I have to decide whether I want to spend my USDC on marketing or on something else. Such opportunity costs make marketing efforts self-limiting. People are not going to give away thousands of dollars left and right for a few likes. Second, marketing expenses don't directly increase revenue unless there's an actual product in the middle that people want to buy and spend money on. I can't spend money to buy attention and immediately make more money on that attention. The real economy doesn't work this way. But in the context of an airdrop, where the tokens are made up and somebody else foots the bill, I can use my token allocation to generate attention to increase my token allocation and round and round it goes.

Attention farming is not airdrop farming in the traditional sense. It is also not typically spam. Some of it presents as desirable content for the platform, increasing overal platform engagement. Nevertheless, I believe it's not desirable when a large fraction of Farcaster traffic turns into attention farming fueled by $DEGEN.

What are the problems? First, it fills up the feed with tons of "do XYZ for $DEGEN" posts. I don't mind those in principle, but I would want to see them only occasionally. When they take over the feed then the site becomes useless for anybody who is not primarily interested in maximizing their $DEGEN tips. Second, in the same vein, the farming posts usually get the most attention in any channel in which they are posted, so it's difficult for other content to rise to the top. This makes Farcaster less attractive for other users to post their content. Why should I put effort into carefully crafting art or blogposts or videos or whatever it is I'm doing when I know my posts won't be visible against the backdrop of attention-farming posts? Third, sufficiently many accounts engage in farming that if you want to have a meaningful $DEGEN allocation for tipping you have to engage in the behavior as well. So again, this is a self-reinforcing mechanism that increases the amount of farming at the expense of other interesting content.

I like the $DEGEN tipping mechanism and I would like to see it continue for some time, but without turning all of Farcaster into a useless morass of attention farming. So let me make two suggestions that are easy to implement and that I believe will lead to a meaningful improvement in tipping culture.

First, the tip allocation curve that determines who receives what allocation should be much flatter. People who are active members of the Farcaster community should not be given a strong incentive to use their $DEGEN tip allocation specifically to farm attention to get a larger allocation. For example, all accounts with the power badge could receive the same allocation. I know this suggestion will get push-back from the accounts that currently benefit the most from high $DEGEN allocations, but I am strongly convinced it's the best way to incentivize interesting content over attention farming.

Second, I think the time window for allocations should be longer. Instead of turning over allocations every day, make them turn over only once week. (Everybody is given their tip allocation at the beginning of the week and then has the full week to give it out.) In part, this would counter-act one issue from my first point which is that nobody would have very large allocations. People who wanted to make larger tips could do so maybe once a week, rather than every day. I think this would be good for many reasons, including the previously mentioned difficulty of having to give away large amounts of $DEGEN on a daily basis. Doing this once a week is much more manageable and can still be very meaningful.

I'm sure this post will be contentious, and people will find all sorts of reasons why what I'm suggesting is bad and a strong culture of attention farming is good. Some may argue that channels can forbid this behavior. (This is correct, and /art does, for example.) I'm happy to listen to these arguments, and I'll keep an open mind, but I find it doubtful that I'll change my perspective. I would much rather see less incentive for farming overall than the majority of channels having to implement specific moderation policies. And I'd also rather see the occasional give away or contest than forbidding these types of posts outright.

First a few disclaimers: This post is US centric. If you're not paying taxes in the US you can stop reading now. And also, I am not a licensed tax adviser or CPA. Don't take anything I write here as tax advice. Everything I say may be wrong. I'm just giving you some suggestions for things to look into and discuss with your tax adviser or accountant. In case of doubt, read the relevant tax law or IRS guidance.

With these disclaimers out of the way, let's start with how the IRS sees airdrops. They see it as income, which you have to declare on your tax return and have to pay ordinary income tax on. Depending on your tax bracket, as of this writing, this could be up to 37% percent in federal tax, plus whatever tax your state demands. California adds another 14.4% in the highest bracket, so if you live in California and make a good income in your day job then over half of your airdop proceeds will go to taxes.

Importantly, the tax bill has to be calculated on the price of the token on the day you claim the airdrop. So, if you claim the token and hold, and then the token price collapses, you will still have to pay the tax bill as if the token had the original value. People have gotten into serious trouble over this, owing way more in taxes than they have liquid capital available. So, before you do anything else, consider whether you have funds set aside to cover this tax liability. If the airdrop amounts to a lot of money compared to the cash you typically have access to then immediately sell enough to offset your tax liability, ideally into USDC so you don't incur any additional taxable events swapping crypto, send the cash to your bank, and then put it into a money market account with 5% interest. In this way, once the tax bill comes due, you'll have the funds at the ready and it won't cause you any headaches. If the airdrop is very substantial you may also have to pay estimated taxes. Here I won't go into this topic, but you should look it up.

Next, you have to think about how you will report this income, as Other Income or as Self Employment. Self Employment incurs an additional tax, but it comes with other perks that are worth it, so I would always try to go the self-employment route if appropriate. You can do this if your business activity is "regular, frequent, and continuous." If you've only got a single airdrop over an entire year then that is probably not the case, but if you routinely farm airdrops and claim $DEGEN every month and maybe do some other things that generate ordinary income then you may very well be able to claim self-employment. I found a good article about the case law in the Journal of Accountancy. And there is also an official IRS publication that compiles all the relevant information for self-employed individuals.

I have heard, but am not sure how true this is, that the IRS is more likely to question income reported as Other Income than as Self Employment. They are concerned about people skirting self-employment tax. In fact, my tax software once gave me a warning that I had too much Other Income reported and that I was going to get audited with this return. I think that if there is any regularity to your activity it's generally safer to claim self-employment.

For the remainder of this post, let's assume you can claim self-employment. This will definitely be the case for any artist or creator who regularly puts out content, receives tips as well as sells NFTs etc., and just generally is generating income on an ongoing basis. (It doesn't have to be a lot of income, it just has to be regular.) I said this comes with perks, so let's discuss the perks. The first perk is the deduction of business expenses. If you're self-employed, you can deduct your business expenses. The IRS writes:

To be deductible, a business expense must be both ordinary and necessary. An ordinary expense is one that is common and accepted in your field of business. A necessary expense is one that is helpful and appropriate for your business. An expense does not have to be indispensable to be considered necessary.

Business expenses are a complicated topic, and people can get (overly) creative in what they deduct and can get in trouble for this, but there are many expenses that you should definitely consider deducting. For example, any expenses you incurred for tax preparation. Do you have to buy the full-service package of TurboTax because your business taxes have gotten complicated? It's a deductible expense next year. Do you need to pay for Koinly to calculate your crypto taxes? Also a deductible expense. And don't forget about gas fees. Did you pay $100 in Ethereum gas to swap your airdrop into USDC and transfer to an exchange? Deductible expense. This is a vast topic and I don't want to go into it in more detail, but it's worth exploring and possibly discussing with a professional tax adviser.

Now the second perk, which I think is even better than the first: A SEP (Simplified Employee Pension) retirement plan. If you are self-employed, you can open a SEP and make contributions to it that you can deduct from your business income. The calculation of how much you can contribute is a bit complicated, and I'll refer you to the relevant IRS documents for detail, but it comes out to roughly 20% of your business income minus expenses or $66k (in 2023), whichever is less. The exact rule is "as much as 25% of your net earnings from self-employment" but those exclude your contribution to the SEP so the calculation is circular. I prefer to think of it as roughly 20% of your income before the SEP contribution. It's important to stay under the limit because if you over-contribute you can create a massive headache for yourself when you have to unwind this position and have to pay taxes on any gains you've made in the interim.

The SEP comes with two perks. First, you won't have to pay any taxes on the airdrop income you contribute to your SEP, and then you also won't have to pay any taxes on gains or trades you make inside your SEP. You will only have to pay taxes once you start withdrawing funds, in retirement. This is a really good deal! If you had a $10k airdrop, you can take about $2k of that and invest into a retirement fund and have it grow tax free. This could ultimately be worth more than the other $8k of the airdrop.

If all the stated requirements apply to you, and you know you're ready to open a SEP, what SEP should you open? How about one where you can buy crypto and make gains and trades tax-free? Well, there are crypto IRAs, for example Alto IRA, where you can do exactly this! A crypto IRA allows you to buy BTC, ETH, and other cryptos and accumulate capital gains tax free. So, if you take the hypothetical $2k from my example, use them to buy blue-chip cryptos in a crypto IRA, and wait a few years, you may end up with a meaningful nest-egg that dwarfs the $8k on which you had to pay income tax. And at a minimum, your crypto investment has to only go up by 3x to compensate you for the taxes you paid on the remainder of the airdrop. (Assume you paid $4k in taxes on the $8k of income, then if your $2k retirement investment goes up by 3x to $6k you've made those taxes back.) I think that's very achievable, and it provides you with some very interesting optionality that you wouldn't have if you didn't have a tax liability from airdrops.

You may have recently joined Farcaster and be frustrated that you don't have an active badge (currently requires 400 followers) and therefore your casts get de-prioritized. Or, you look at friends who joined just a few months earlier and they have thousands of followers while you have barely breached a hundred. So, now you want to remedy this injustice as quickly as possible. You want thousands of followers also. Ideally tomorrow.

Now, first and foremost, take a deep breath. You have plenty of time. You can't expect to build a meaningful following in a week or two, but if you take a longer-term view (and in my mind, that means a couple of years at minimum) you can build a strong audience that will be the envy of new Farcaster users a few years down the line. I will discuss some strategies here that will help put you on this trajectory, but the most important strategy of all is sticking it out. Most people don't have the patience, consistency, or stamina to execute a long-term growth strategy. If you can commit and stick to a strategy you're already ahead of 99% of the other people on the platform.

Next, your goal may be to get as many followers as possible, but I don't think that's a useful goal. You want quality followers, followers that give you engagement when you cast. It's better to have 1000 followers who actually care about what you're casting than 50,000 who don't. I have first-hand experience with this from Twitter, from before Elon took over. (To be clear: this is not related to the broadly lamented decline in engagement people have experienced there over the last fifteen months.) I had about 20,000 followers, mostly from my work in data science, data visualization, and R package development, and when I shifted my interests to web3 and generative art many of those followers just didn't care. My engagement numbers for tweets on web3 topics were comparable to those of generative artists with 500–1000 followers. Alignment between your followers and what you're posting is extremely important. I didn't lose my followers on Twitter (I still have about 20,000) but they stopped engaging.

As of this writing, I have almost 4000 followers on Farcaster, and I only joined three months ago. So I'd say I've been reasonably successful in building an audience from scratch. Now to be fair, I benefitted tremendously from the release of frames, which led to a massive inflow of new Farcaster users and caused everybody's follower counts to go vertical. However, even before frames I was gaining about 200 followers a month, which is very respectable. And, since Farcaster is still so young and small, there will be other events similar to frames that will cause further spikes in follower counts for active users. You may have missed the frames opportunity, but you don't have to miss the next one. Be ready.

Profile page and bio

Let's start with your profile page. First, make sure you have an engaging bio. Why should I be interested in you? Put something there that indicates why you may have a valuable voice in this space. Maybe something about what you can do. "Smart-contract engineer." "Generative artist." "Professional chef." Whatever it may be. Or something about your work. "Founder at XYZ." "VC at ABC." Whatever it is, make it stand out and unique. If you write "I like crypto" or "Hi" or "To the moon" nobody will care. Also, don't write "Will follow back" or "follow 4 follow". Why should I follow you if the only thing you've got to offer is you'll follow me back. You're almost certainly not going to be a quality follower for me.

Then, add some links to other sites where you have a presence. If you've got a linktree, definitely add that. Otherwise, link to your personal website, or your Twitter account, Instagram, or NFT gallery. All these links provide potential followers the opportunity to learn more about you and feel more confident that following you is a good idea.

Put some effort into your profile picture. For sure, not setting a profile picture at all is a bad idea. The profile picture helps other casters to remember you, so it needs to be unique. The default picture is not. If you've got an established brand or logo, use that. (For example, I always use the same image everywhere.) Alternatively, if you have an NFT from a famous collection and identify with the vibe of that collection, consider using that (e.g., a punk, or an mfer, or an opepen). But really, you can use any image that looks distinct and works when scaled down to a small size. Just be careful with images that have tons of detail. When scaled down, they may look like mush. Mush is not memorable.

Finally, make sure you've got some interesting casts in your timeline. In particular, if your main "Casts" tab is completely empty I may have doubts about who you are and whether I should follow you.

Casting

Now that your bio is done, let's talk about casting. If you're coming from a platform that is highly algorithmically curated, such as Instagram, TikTok, or Twitter, your approach may be to cast as much as possible and try to get your casts go viral. This strategy is unlikely going to work on Farcaster. There is no strong algorithmic curation at this time, and for sure no algorithm that tries to maximize engagement. The algorithm on Warpcast (the default client) mostly just tries to hide the noise. Chances are your posts will look like noise. So this is a losing proposition.

Farcaster also doesn't have hashtags, but it does have channels. Channels work more like subreddits on Reddit than like hashtags on Instagram or Twitter. You have to actively cast into channels, and each cast can only go into one single channel. Tagging channels in your cast has no effect.

You should definitely cast into channels. This will increase the visibility of your cast, in particular if the channel is not extremely large and busy. Your cast will be visible to everybody subscribed to the channel, which likely is more people than are following you (if not, why are you reading this post?), and as long as your cast is on-topic for the channel all these people should be interested in what you have to say. Pick a handful of channels and keep casting into them repeatedly, rather than spraying your content all over, so that the followers of those channels start recognizing your name.

While casting regularly into the same channels is extremely valuable, make sure you don't overdo it. If you cast the same thing five times a day into the same channel people will quickly get tired of you. My rule of thumb is almost everybody can cast once a day and not annoy people. If you want to cast much more than that, say more often than three times a day into the same channel, you have to be really good and have tons of diverse content. If you don't, just don't do it. If you're an artist and you want to sell that NFT collection, don't post about it more than once a day into the same channel. However, posting one image a day consistently over a long period of time is a really good strategy. It shows consistency without getting on people's nerves.

Most importantly, engage with others on the platform. Don't just cast your own stuff. Instead, engage and interact. While likes and recasts are engagements also, and are valuable, they don't replace actual replies. Ask questions. Make witty comments. Provide answers or counterpoints. Engage. And if others engage with your casts, engage back. If somebody makes a witty comment, give a witty response. If somebody asks a question, give an answer. If somebody challenges you, defend your position or acknowledge they are right. Say something.

Another good growth strategy is transferring content from one channel to another. Find two channels that are related, say /food and /pasta, and whenever you find a banger cast in one channel just quote-cast it into the other, for example saying "Found in /food." This allows you to get some credit for banger casts that you would not have come up with yourself, and it also advertises your account to the person who you're quoting. Double-win. Of course this strategy extends beyond just multiple channels on Farcaster. You can also repost content from Twitter, or from Reddit, or from the New York Times. Whatever works for you. Just make sure you keep casting quality content.

Finally, there are traditional growth strategies where you give away NFTs or other rewards for follows, recasts, and/or likes, and these techniques definitely work to get your numbers up. However, you're unlikely to gain quality followers this way, so use these strategies carefully if at all. You don't want to get a reputation for just constantly engaging in these transparent and shilly growth strategies. The real quality followers will likely be put off and quietly disappear.

Following

It's important to follow others on the platform. You can't expect to be followed without following yourself. If you're too stingy with following others you will likely miss out on a lot of positive interactions and opportunities to grow your network.

At the same time, don't just follow everybody. I always get skeptical when I see accounts that follow thousands of accounts while having relatively low follower counts themselves. This suggests low quality control. Why would I want to follow somebody who doesn't seem to put a lot of effort into curating their own feed?

There are two types of accounts you should follow. First, follow the biggest and most visible accounts on the platform, in particular in your area of interest. Those are the thought leaders. They set the agenda. Everybody reads their casts. So you should too. And, their casts should give you plenty of opportunity to reply. As much as possible, when the big accounts talk about something, try to insert yourself into the conversation and make (intelligent, thoughtful) comments and contributions. Even if the big accounts don't care about what you have to say, their followers may start getting to know you and may start taking you seriously. And at a minimum, you know what's going on on the platform, what are the issues and trends, and this should inform your own top-level casts. Cast things that speak to the current topics of debate.

Second, follow active accounts that are relatively small and don't have a lot of followers yet. These accounts are likely in the same boat as you are and you guys can help each other out. But don't do it with "follow me I'll follow you." Just follow them. My rule of thumb is: If I've had a few good interactions with somebody (possibly even just one), I'll just follow them then and there. More often than not, they'll follow me back. And if they don't, no harm done. If they're interesting I'll encounter them again and interact with them again and maybe they'll follow me in the future. And if it turns out they're not interesting and their casts start to annoy me I can just unfollow then.

Don't discount how effective of an advertising mechanism a follow can be. For smaller accounts, they likely monitor everybody that follows them. So, go through the channels you are interested in, identify people with interests similar to yours, and follow them. However, and this is super important, only do this after you have completed all the steps under "Profile page and bio." If you've got a nondescript bio your follow won't be a good advertisement for yourself. On the flip side, if your bio rocks then every time you follow someone with aligned interests they may likely follow you back.

Channels

Finally, it may be a good idea to start a channel. A channel can give you additional visibility, as people may be interested in the topic of your channel but would otherwise not have thought of following you. You may think that all the interesting channels exist already, but that's almost certainly not the case. Look at how many subreddits there are on Reddit. And people keep still making them. Imagine traveling back in time to Reddit in 2007 and saying "I'm late, all the interesting subreddits exist already." Well, I can tell you for certain there was no cryptocurrency subreddit, let alone any of the more specialized subreddits such as bitcoin, ethereum, solana, tezos, etc. There wasn't even any cryptocurrency in 2007. So, if you want to organize a channel, you can almost certainly find a topic that isn't yet covered. You may just have to go a bit deeper and find your unique niche. For example, maybe there's a /food channel and a /pasta channel but no /gluten-free channel. Or maybe there is, but no /gluten-free-pasta. And so on. Find the niche that appeals to you and that has been overlooked.

Now, let me be frank, running a channel is a commitment. If you're not up to it don't do it. There's nothing worse than a half-abandoned channel that looks like nobody cares. Or a channel that is overrun with spam because moderation has broken down. If you start a channel you have to be prepared to be the person who drives it forward, who posts at least once a day, who encourages other people to contribute, and so on. On the flip side, the worst that can happen is it's a failed experiment and eventually you abandon the channel and don't renew it (remember, channels have a yearly fee), so why not. Give it a try.