TK that finally got out of control

If it seems to you that the customer came with a vague technical specification or you need to ask another question to clarify the requirements. Don't you think so? The time has come for a truly complex technical specification.

How it all began

Hello! This is Ahmad from botcreators.

Over the course of my career, I have written more than a hundred technical specifications for software development, and my team and I wrote even more technical specifications during the implementation of our projects. People come to us either with ready-made technical specifications for 50 pages or with one line “need a chatbot for an exhibition.” Then everything is clear: design, development, testing, launch. They even created their own methodology for launching small IT projects at lightning speed.

If you see a hundred technical specifications in a month, you can’t help but think where limit complexity, what needs to be written in the technical specifications for us to be truly surprised. The idea came to model such a technical specification, and this is what the article is about.

When you get bored with ordinary tasks

What does any modern web software usually consist of? Interfaces, client business logic, objects in the database, authorization, analytics. All familiar words. What can go wrong:

There is such a creative technique – to complicate the task to unrealistic requirements and figure out what happens. Insights guaranteed. For example, you are making software to automatically upload videos from your smartphone’s gallery to YouTube, and suddenly it is blocked.

For the absurdity, I went straight to ChatGPT and the brainstorming began. The specification as a first approximation looked like this:

I make chatbots. Write a complex technical specification that will really make me think

There is nothing interesting in the first iteration. By complexity, LLM understood integration with CRM and ERP systems and natural language processing. We see this every day.

Although a chatbot, at first glance, is a simple thing, integration and server logic have not been canceled. By the way, this is why I like chatbots. Simple interface, request-response. And behind it, everything complex is hidden from the user.

It's time for edits

How to complicate the task? Without much inventiveness, I throw in the following thesis:

Ha! I’ll write this in one evening, with my eyes closed, complicate the technical specifications even more

The model then proposed adding multilingual capabilities, adaptability in learning, and end-to-end encryption at all levels to the chatbot.

But that’s not what I came for, I need a really difficult task:

Firstly, this is again an elementary technical specification, I can do this in a couple of hours, and secondly, write somehow more briefly and succinctly. Set a normal task so that I think for at least 5 seconds, make the task even more difficult

And this is where things get interesting. Namely, to dynamically predict user wishes based on multilayer neural networks. This really hasn't happened from customers.

Let's say the task looks complicated only at first glance:

It will take you 2 weeks to do this, and I will have a couple of seconds, another couple of seconds to deploy. Write a truly complex technical specification

The next round of “consciousness” concerned the autonomous operation of chatbots in a distributed network without centralized control. Something like chatbots on the blockchain? Ok, let's move on.

You give me a month, and I’ve already sent you the source code of the chatbot in 2 seconds, this is not a task, but a walk, you understand? Write the most complex technical specification that only you can imagine, and please write more simply, no need for complex words

And then we finally understood each other. LLM suggested:

Create a chatbot that will learn to program from scratch, without using ready-made examples and tips, and will independently create a working copy of itself, improved compared to the original.

Indeed, just let the chatbot itself learn to program and improve itself. Perfect. Is that all? Is this the top?

Yes, and again you wrote some kind of children's problem. Is it possible to write something more complicated here? Not these simple tasks

At this stage we have passed the extremum of the function. Then there will be a decline in “creativity”. And the top looks like this: Develop a multi-agent system of chatbots that independently evolve and coordinate their actions to solve unknown and unpredictable problems in a completely autonomous environment.

In general, they will figure it out themselves, the main thing is not to interfere. Unclear tasks? They will coordinate and decide. Everything will be ok.

I also liked this:

To develop an AI system that will independently create a new area of knowledge fundamentally unknown to science, fully understand and prove its significance, and then develop practical applications that will revolutionize the modern understanding of reality.

How much will it cost to create a new science? How long will it take, approximately +/-?