The Secret Chatend Master Plan

Andi, Founder • July 22, 2023

Over the past few months I have been working on the company known as Chatend. As you know, Chatend is a bespoke chatbot development platform as well as a chatbot consultancy. Here I will put forward my vision for the company and explain why I think Chatend is a platform worth building on.

Why chatbots?

I have been working in the field of web development for over 13 years now. Once I taught myself HTML, I started hosting with cPanel and PHP and slowly moved my way through node.js, React, and Next. Since then I've designed database access patterns and rolled out serverless APIs. I've built websites, web apps, and mobile apps. I've worked with a lot of different technologies and teams but there are a few consistent trends with software ideas:

  • Many ideas will never see the light of day because hiring a development team is expensive.
  • Every funded idea is behind schedule and over budget.
  • Every idea is built with a mixture of technologies that felt right at the time, but soon show their incompatibility with frequent use.
  • Every idea simultaneously has a frontend that is too complex, but not feature rich enough.
  • Every idea uses similar interface patterns (ex. passwords, navbars, Like buttons, etc.) to familiarize users with the app.
  • Every idea inevitably needs the same 4 backend patterns: Create, Read, Update, and Delete (CRUD).
    • Sometimes it feels like all we do as developers is build CRUD apps.

In November 2022, when I started building AI chatbots with GPT-3 for the first time I realized there was an opportunity to dramatically shift the way we build ideas. It started with the question:

When we achieve Artifical General Intelligence (AGI), the problem of software development will be solved instantly. AGI will simply develop all the software we ever need. But how does this actually look like? How do we communicate to the AGI what we want built? And what does the end product look like? <br/><br/> What does it actually build?

Chatend is the result of working backwards from that question.

Here's the bet: AGI won't start a new software project by reaching for Next.js or Qt or Electron. It won't start by building a web app, or a mobile app, or a desktop app. It will start with a chatbot. And it will start with a chatbot because chat interfaces are the most universally understood user interaction paradigm of all time. Everybody has texted before. Everybody knows how to use a chat app. And so, everybody knows how to use a chatbot. ChatGPT's wild success proves that people are ready to interact with chatbots on a daily basis. The problem with chatbots of the past was not that chatbots are a bad paradigm, but that the technology wasn't good enough. The conversational experience sucked. With Large Language Models, this is no longer the case.

So, for every software idea, Chatend starts with a chat interface. We certainly don't end there - we can add tables and fancy graphics and mobile apps later - but we start with chat. This solves a ridiculously long list of concerns up front, with some of the most important being:

  • Where do I host my idea?
  • What does the idea look like?
  • How do users access my idea?
  • Do I have to make people download and install something?
  • How do I make my idea accessible to users with disabilities?
  • How do I teach my users how to use my idea?
  • What's my frontend/backend stack?

We host your idea on Chatend. The idea looks like a chatbot. Users access your idea from any existing chat app they already have installed like Slack or Discord. Because it's accessed via a chat app, it's already cross platform and accessible from desktop, mobile, and tablets. Users already know how to interact with your idea: just text it as you would a person. Users immediately gain access to information and can start using your idea as soon as they learn about it. And your stack? Well, Chatend can do it all.

The next step is agents: chatbots that can carry out actions on behalf of their users. Chatend solves this with Custom Commands. Custom Commands allow you to write and deploy custom TypeScript code to your Chatend bot. This code can be used to carry out actions on behalf of the user. For example, you can use Custom Commands to create a new user in your database, or send an email, or create a new record in your CRM. Custom Commands are the key to building chatbots that can do anything. Of course, running code from a bot is not a new idea, but the breakthrough technology that allows this to go mainstream is GPT's ability to understand and respond to user input. Concretely, it means we can present a GPT model with an existing API - say, commands and their parameters - and teach it to execute commands from that API with the proper arguments. Users are free to speak in whatever way they want and trust that the proper command will be executed with the right arguments. This is a huge step forward in the chatbot space.

You may wonder how this is different from ChatGPT Plugins. The problem with Plugins is that they are hosted off platform. If you want to build your own Plugin, you need to host it on your own server and deal with all the administrative headaches that come from that. If you want to use someone else's Plugin, you need to trust that they won't steal your data. Chatend solves both of these problems by hosting your Custom Commands for you. Additionally, Chatend will provide a marketplace for Custom Commands so that you can easily find and install new Commands for your bot.

Finally, data. Chatend Keystore is a first party key-value store that allow Custom Commands to save and retrieve JSON data. It requires no setup to use and easily integrates with 3rd parties using a REST API. Additionally, we'll support the AWS node.js module to allow your Custom Command to interact with all AWS resources including databases.

The end result of all of this is Chatend IDE: a one stop shop to build, deploy, and manage your chatbot. You'll be able to create Custom Commands and watch how your bot's behavior changes in real time. You'll be able to manage your deployments and see how your bot is performing. You'll be able to test your bot with a live chat interface and see how it responds to user input. And you'll be able to do all of this without ever leaving the Chatend IDE.

Why Not No-Code / Drag and Drop?

Why not just build a no-code interface so that users can drag and drop their way to a chatbot?

There have been millions of opinion articles written on this already, so I will keep the argument short. The problem with programming is never learning the language or remembering the syntax. The problem with programming is that code is law and thinking like a lawyer is difficult. It requires a lot of mental energy and background context. Accordingly, most no-code tools end up being used by programmers rather than non-technical people. Programmers already know how to build what the tool is offering, so they use the no-code tool to save time. When they hit a wall or a platform limitation, they switch back to code. This is why no-code tools are so popular with programmers: they provide a great foundation and help avoid boilerplate. But they are not a replacement for code. That's why Custom Commands accept custom code: I want Chatend to have as few limitations as possible. When you need to give your bot more powers, I don't want you to reach for an AWS server or a Lambda function. I want you to write code on the Chatend platform directly. I want you to be able to do anything.

The Business Plan

Chatend will start out doing bespoke chatbot engagements. I'll onboard customers one by one and ensure Chatend's technology performs well under pressure. Reliability is critical, and I need to get it right. I'll use the revenue from these engagements to fund more development of the Chatend platform. Once finishing touches are made, I'll also release Chatend IDE for anyone who wants to be able to sign up and create their own chatbot.

This leads up to Chatend Assistant, which is the ultimate goal of Chatend: a bot, that you talk to, that builds other bots. This goes back to our AGI hypothetical. What would it look like if you could build software by talking to it? With Chatend complete, there's now a way to host bots, their data, and their commands from one central platform. We can build a bot that talks to Chatend IDE and does everything you need to configure your new bot appropriately. In the future, you won't even need to know how to use Chatend IDE, you'll just talk to Chatend Assistant and it will build your bot for you.

With Chatend Assistant, we can go further beyond chatbots too. Think about what you've accomplished by building a Chatend bot: you've built out a set of functions and data access patterns. Chatend can use this information to build Web and mobile frontends around your bot. Imagine waking up one day and having an entire app built for your idea already. And as LLM technology progresses, your idea will get better and better over time with less and less effort on your part.

This is the future. I want to build it. Let's bring it here together.

- Andi