Recently, I watched an interview where Sam Altman said that their 03 model is currently the 175th best competitive programmer in the world. Their internal benchmark is around 50, and they aim to reach number 1 by the end of 2025.
Then we have DeepSeek, which used reinforcement learning, was built for under $6 million, and shocked the world with its insane capabilities in coding, math, and more.
This challenges the idea that an LLM model needs billions to train.
On top of that, we have AI-powered tools like GitHub Copilot, Cursor AI, CodeRabbit AI, and others that are making the coding process even faster.
These advancements prove that the way we write code is going to change.
In short, AI will eliminate the need to manually write code, and nearly 90% of programming jobs will be automated.
What's next?
Programmers will shift their focus to solving real problems and building groundbreaking applications rather than just writing tons of lines of code.
And that's what I'll be discussing in this post - along with evidence, recent developments, and why programmers need to adapt.
Let's start.
Building software used to be hard
As a programmer, when I try to build software or even a website, I need to follow a traditional process.
In the programming world, we call this process as SDLC (Software Development Life Cycle).
And the SDLC process looks something like this:
You see, it has multiple steps, and the middle ones take the most time - sometimes months or even years to implement.
Even then, the software can fail to run due to mistakes, heavy traffic, or other unforeseen issues.
If you look at the SDLC process from a broader perspective, you'll realize that most of the focus is on inputs, and only after a long time do we see an output - which then requires further improvements.
But that's not the end.
The team behind the company also has to focus on marketing, adding trending features, finding strategies to increase revenue, and much more.
You see, the process was actually difficult.
And if you weren't a programmer, forget about SDLC or creating software - it simply wasn't for you.
But now, AI is changing everything.
AI is rewriting the rules of software development
Now, you don't need to be a skilled programmer to write programs.
We have LLMs like ChatGPT, Claude, and DeepSeek that only require a specific prompt about what you want to create - that's all.
If you spend a few more hours making tweaks, your MVP (Minimum Viable Product) is ready.
I wrote a post earlier where I shared how you can build your dream SaaS product, custom dashboards, complex apps, and more using LLMs like Claude.
In short, with LLMs, anyone can create whatever app they want using simple prompts.
And this is just the beginning. Yes, you read that right.
Over time, LLMs will become even better and smarter - so much so that 90% of programmers who are just writing code will eventually disappear.
The rise of AI-powered tools
Now, this is just the basics, and most of us are familiar with these foundational models.
But there are even multiple AI-powered applications that are built on LLM models like ChatGPT.
Let me be specific and take the example of Cursor AI - it's an AI code editor, and one can use it without any coding knowledge, just by prompting.
Here are some great features it offers:
It also has an agent mode that can handle tasks from start to finish.
And it can even automatically write and run terminal commands, find coding errors, and fix them.
There are also AI-powered code reviewers like CodeRabbit AI, which work with your pull requests to provide automated code reviews, guides, and more.
Yes, it integrates directly into your pull request, and the code review process becomes automated.
And the shocking part? Most of these AI-powered tools are making millions in revenue with just a few employees.
Here's a tweet to prove my point:
Insane, right?
The world's first AI software engineer
Now, let me tell you - AI-powered applications are just the beginning - we're now seeing AI agents evolving at an even faster pace.
Devin AI is a great example, and the team behind it called it the first AI software engineer.
When it was initially released, the official blog mentioned that Devin can:
- Learn how to use unfamiliar technologies
- Build and deploy apps end to end
- Autonomously find and fix bugs in codebases
- Train and fine-tune its own AI models
- And many more such features
But at that time, programmers were a bit skeptical since the company admitted that Devin could solve only 13.86% of issues end-to-end.
However, within just six months, they released a new post stating that they had achieved 34.6% using o1-mini and 51.8% using o1-preview.
Insane, right?
Now, I don't want you to be worried, but this is just the start. We are seeing more AI development, better LLMs, and even more advanced AI agents.
And as competition increases, it's only going to accelerate further.
What's the future?
Well, no one can actually predict the future.
But based on current research and what the AI experts are focusing on, it's clear that the next few years will see a rise in AI coding assistants and fully-fledged AI agents.
Even top companies like OpenAI, Nvidia, and Microsoft are heavily investing in building these AI agents.
But what exactly are AI agents?
Think of them as AI-powered "employees" that can take on complex tasks autonomously. These agents will be capable of completing entire processes - 100% - without requiring any human intervention.
In short, tools like Devin AI will become more precise, with issues like hallucinations virtually disappearing.
And I believe that the programming field will be one of the first to fully adopt AI agents.
As a result, traditional manual coding will likely disappear, with programmers shifting their focus to problem-solving and higher-level design tasks.
My key takeaway
Now let me be honest, and get straight to the point - it doesn't matter to customers whether you're using AI or not when building your software.
They want features that directly solves their pain points, or streamline their workflows by providing effective, time-saving solutions.
And if another app, built completely using AI, provides better solutions, your customers will switch to it.
The only solution left is to start integrating AI into your programming journey.
Begin by using AI-powered tools to automate repetitive tasks, reduce manual coding, and make yourself more productive.
And if possible, begin learning AI so you can build custom AI agents specific to your needs and more.
For programmers, learning AI is much easier since we already know how to write code, understand programming languages like Python, and have trained our minds to think through logic and problem-solving.
Of course, make sure not to share sensitive information like API keys with large language models (LLMs) like DeepSeek, but other than that, feel free to use AI.
Lastly, whether we like it or not, AI will keep improving, and eventually, everyone will be using it, regardless of what they believe now.
Hope you like it.
That's it - thanks.
If you've found this post helpful, make sure to subscribe to my newsletter, AI Made Simple where I dive deeper into practical AI strategies for everyday people.
Top comments (3)
Ai is an industry disrupter never seen before. Its all encompassing, can be everywhere and will only improve. Would love to see it totally replace scrum masters and managers. In fact we don't need government any longer. We just replace with an answer review board.
how we can protect our jobs against this
It might be a good idea to learn how to use it and get better. AI is really good as a complementary tool.