How Discord and AI will shape Web 4.0

A couple years ago, I was going to publish an article how crypto exchanges were going to be replaced by wallets. The main premise is: If exchanges were to be decentralized, it makes no sense to use a centralized exchange. And since you don't need to login or even visit the DEX, as you can execute the trade directly from your wallet, then wallets will become the main gateway for 99% of users. Users won't know what exchange or liquidity provider is executing their order; and as long as it works, they won't care. This will sunset the business of exchanges as we currently have it.

While wallets are as important as ever, this didn't fully materialize. Mainly because the industry still rely on multiple centralized exchanges, and it doesn't seem that they'll be replaced soon. That being said, I still think this will play out on the longer-term, though maybe a little bit differently.

See, I've been a bit of a tech hermit in the last few years as I've been alienated with the constantly degrading web. It has been almost 4 years since I wrote that article; and the web only kept getting worse. It has become an excruciating pain to get anything done on big tech platforms. This, being mostly driven by companies trying to extract as much as possible from their current audience. Which is, not surprisingly, what dying companies and business models do.

So what will replace the current incumbents?

Discord, Telegram, AI, TikTok; or some mutant of these.

I hate TikTok but I have been recently introduced to Discord and Telegram (I am old now); and while I hated the concept of gating and the severe lack of openness, something clicked. Discord and Telegram have a bit of a learning curve, but it's not too high and it pays off handsomely.

You might be wondering how platforms like Discord or Telegram, originally created to form communities, could replace the web. These platforms were made initially for discussion. However, they allow a great deal of flexibility inside them. Your service can integrate with them through "bots". This allow the users of the platform to interact with your service through the platform UI.

Suddenly, you don't need a UI or a front-end. Or more specifically, you don't need an HTML/JavaScript UI. You use or integrate the UI provided by the platform. This is beneficial for users since they are already accustomed to the platform interface. They don't need to re-learn a different UI.

This might sound like a bad idea, but MidJourney recently launched exclusively through Discord. And it worked well. In fact, the additional social aspect was quite beneficial. On-boarding is a breeze too, there are no user sign-ups!

Constraint as a feature

You might be thinking that the UI constraints of these platforms is an issue. But I think of it as a feature instead. Most users struggle through the average UI: They have to learn, remember, click buttons, and probably deal with some bugs here and there. They might want to adjust the size of the font, the size of their window, or maybe they have special preferences.

HTML tried to make standards for these issues but developers seldom respect them. Most UIs out there are terrible, and worse, they keep changing. Users don't want to learn a UI or learn about your new color scheme. They want to get things done, and fast.


If you can't architecture complicated interfaces, you'll need to rely heavily on text to interpret user intent. You can program commands that the user can execute, but users can only learn and remember so much. You need a process to interpret user intent.

This is where AI comes in. Users will interact with your application with natural language. They'll type commands in plain English, and your process will analyze this text to determine what to do.

Virtual Reality

No. Not a single chance. VR is the complete opposite of what we described above. The UI in virtual reality is at least as complex as the 2D HTML/JavaScript one. And this is the best case. This is not an opinion against VR headsets. I could see myself using one; for watching a movie, playing a game, or using it as a regular screen.

However, VR adds a ton of complexity to interfaces. This will make interfaces worse, not better. Paradoxically, the best interface (command-line) was introduced in the 60s; and it still is the case today. It is the least taxing to users: you just type a few words.


While we are not quite there yet, there is already a few working examples: Many Generative AI companies are opting to use Discord/Telegram as their interface. You can now execute (crypto) trades through Telegram. This is quite niche and limited but it's still the early days; and AI is still quite limited, expensive and closed (GPT-4)

The future

In the future, you'll login to your favorite platform. You have already connected your bank, brokerage, email, storage service, etc.. to this platform. You'll visit your bank branch (ie: group) in the platform. You'll be able to type commands like: balance show or naturally ask for There is a fraudulent transaction in my credit card in the 2nd of last month, or order a new card to my platform address.

If you are doing generative AI, you don't need to download the file from your cloud storage provider to your computer, then upload it again. You'll be able to transfer it through the platform itself. Quite handy for big files.

Your friends will be in the platform as well. Your company dashboard too. You don't need a different one (ie: Slack) because the current platform can support whatever functionality Slack can do. So you'll naturally be there too.

This is the Virtual World of Web 4.0. But instead of it being an immersion of 3D artifacts in a VR headset, it's just a chat box.

Published on

Previous Article

Subscribe by email

Get new publications from this blog directly to your email.