Developers Prefers AI Chatbots for StackOverflow

69% of users prefers AI chatbots according to IBM. Figures like this doesn't surprise us. An AI chatbot can answer support requests in seconds, while creating a manual support ticket implies waiting for days before you get an answer.

StackOverflow fundamentally is a ticket system, with its only difference being its answers are publicly available. So the same figures (probably) applies to StackOverflow too.

StackOverflow was Brilliant

For a software developer like me, StackOverflow used to be brilliant. For large open source projects it gave me an answer in seconds. A Google search typically showed me a StackOverflow answer in seconds, and I could move onwards to my next problem within 30 minutes.

However, if you were a proprietary company with a proprietary software platform, you couldn't use StackOverflow. And even if you could, chances are that you had too few people interested in asking questions to generate a reasonable amount of answers to support future clients. StackOverflow basically didn't work for companies with a proprietary software system. This was by design because they didn't want you to use StackOverflow to promote your product.

Even for larger open source projects StackOverflow sometimes didn't work, because they didn't allow you to use it for community guidelines reasons. I've been operating and managing open source projects with millions of downloads where I couldn't use StackOverflow because, and I quote; "was promoting my platform."

AI is better than StackOverflow

StackOverflow's traits are as follows:

1. Users can ask questions
2. The community can answer these questions
3. Crowd sourcing ensures that 95% of the time the best answer "bubbles" to the top
4. StackOverflow works in a symbiotic relationship with Google, allowing you to rapidly search for solutions using Google, having StackOverflow provide the answer

If you break down the above process into value propositions, and you ask yourself how an AI chatbot can deliver the above better, you'd get something like the following:

1. You can ask the AI chatbot a question
2. The AI chatbot answers the question using its RAG database as "context"
3. Quality assuring answers during deployment ensures high quality answers where 95% of the answers are accurate
4. Because you're using RAG and VSS, the AI chatbot has "integrated search" for relevant information, hidden beneath its surface, completely replacing Google in the above equation

By combining AI with your existing documentation to create a RAG database, you've basically created something that's 1,000 times better than StackOverflow. And you control it yourself.

How an AI chatbot works

An AI chatbot is actually quite easy to understand. Most AI chatbots are based upon RAG. RAG is an acronym and implies "retrieval-augmented generation". It works by having a database of knowledge typically based upon your system's documentation. This database contains knowledge about your system.

When a question is asked, the AI chatbot will perform vector similarity search into this knowledge base, and find the top 5 records most relevant to the question being asked. This process replaces the search parts from StackOverflow with VSS, allowing each question to find snippets of information from your knowledge base somehow relevant to the question being asked.

The AI chatbot will transmit the relevant parts of your knowledge base to the LLM, together with the question, and instruct the LLM to answer the question using the context provided. The LLM itself will extrapolate missing information and use the context to ensure it provides you a correct answer, even without complete information.

Logically this becomes the equivalent of you reading hundreds of pages of text in your system's documentation, for then to semantically understand each page, and generate an answer for the question based upon your system's documentation. The difference being that the AI chatbot does this in seconds, while you'd need hours to do the same.

This process is actually so fast and accurate that even I use our own AI chatbot on Magic Cloud to help me during coding. As the architect of Magic Cloud I'm obviously the authority on the platform. However, our own internal AI chatbot can still find the answer so fast and produce such high quality that even I prefer using it for the code I wrote myself. Because there's parts of Magic I haven't used in months, maybe even years. Over time I forget how to use this code. As I forget the details about it, I can ask our AI chatbot a question about Magic, and the AI chatbot gives me a high quality answer in some few seconds. See the screenshot below for an example.

You can try it out for yourself here.

AI roadkill

Since ChatGPT became popular 18 months ago, traffic to StackOverflow has dropped like a rock in water. They've lost 50% of their traffic the last 18 months, and had to lay off 28% of their staff as a consequence. StackOverflow has basically become "AI roadkill". The reason is because most software developers have replaced StackOverflow entirely with ChatGPT, Microsoft CoPilot, and Claude.

If I search for a solution to my problem I have to wade through 15 ads, then click the StackOverflow link, read the question to see if it's relevant to my problem, for then to find the most accurate answer further down on the page. Then I have to transpile the solution to my own code to make sure it works according to my structure. This process takes me 30 minutes.

Using our own AI chatbot, I can instruct it to generate code according to my wish. I can even give it my own code's structure and tell it to modify my code to add new functionality. This allows me to take a process that would previously require 30 minutes and optimise it down to 3 minutes. 30 seconds to write my question, 30 seconds to have an answer, and 2 minutes to proof read its resulting code.

For me personally AI has optimised the way I work by 10x

The future of StackOverflow

For more than a decade StackOverflow was like crack cocain for developers. When you were stuck with a problem, you could have StackOverflow provide an answer rapidly. However, you could not use it as a business, you could not use it to promote your own software, and you could not generate business intelligence from it. This made it useless for your company.

To further the insult, StackOverflow would do everything it could to provide "new job opportunities" to your developers, by adding job ads to their answers - Resulting in that from a business perspective it was poisonous. One of your devs being stuck at some problem in MimeKit and 5 minutes later he's leaving your company for "a better opportunity" at some other company.

AI chatbots fundamentally changed this equation, giving you control over your own documentation and help systems, in addition to providing you with better customer support.

Wrapping up

For a software company delivering software solutions to their clients, AI chatbots are simply superior to StackOverflow. In addition, studies shows us that 69% of your users already prefers AI chatbots instead of StackOverflow and other ticket systems. The reasons are easy to understand; The AI can answer any question, and it can answer the question 1,000 times faster.

This implies that you can use AI chatbots for your software company to deliver customer support, outperforming StackOverflow, while saving thousands of dollars each month.

For StackOverflow of course this is sad, but after having StackOverflow "shadow banning" Hyperlambda and Magic Cloud for 15 years, I'm perfectly fine with that. They became arrogant over time, making it increasingly difficult to gain access to their "community of software developers." In addition they became poisonous with time, and stopped innovating.

Today there are no reasons to chose StackOverflow, ignoring the fact that 99% of all software out there cannot even chose it even if they wanted to. Today AI is simply superior. AI's destiny is to change the way we work by optimising our processes. As it does, a lot of our existing ways to work will be negatively impacted. Google is struggling to find their way in this new world, and so is StackOverflow. However, that's because fundamentally what we needed wasn't search or generic answers. What we needed was personalised solutions to our own problems. What we wanted was something allowing us to discuss our personal problems, and not some generic "one size fits all" solutions - And AI is simply superior in delivering that.

Goodbye StackOverflow, Hello AI