This post is for students who like programming but are uncertain of its future. I am not very good with arguments, but I have been in this business since 2000 and people often ask me about what the future holds, no one can know what that will be but we can speculate with past experiences. So, this is my take on programming in 2025, let me know in the comments what you think.
The first concern for every programmer is AI. AI-generated code will not make programming obsolete, nor will it take coding jobs, though it will/has started to morph how we program.
Before we start, LLMs at the scale the world is getting used to, are very resource-hungry, and there are concerns like where they get their data from and what they do with the data provided as input.
LLMs do not reason as humans, they are statistical models that learn patterns and can generate outputs based on those patterns. Creating these models is expensive, and even the reinforcement learning-based AI models that train on other models’ results are not cheap.
LLMS is a simile to calculators which never made teaching and learning calculations obsolete, rather we teach our children mathematics with other skills. We learn to drive yet the future is of automated cars because it is our present need, we might not need drivers in the far away reality, but that is ok, we need this skill just like basic calculations. We need to know how to reason to take advantage of automated reasoning.
2nd argument is we need to understand what the code AI generates and whether it fits our purpose, as these AI tools advance we can expect more refined output, but for the foreseeable future human input and audit of the final product are necessary, especially important for industries where human life is at stake.
I have drifted from programming to just reasoning in the last 2 arguments, in my defense programming is reasoning using a programming language, getting your computer to do what you want with logic and instructions.
3rd these LLMs require researchers, programmers, and a support system, and a demand for new, more powerful hardware means that software will need to adapt to its requirements. The complete ecosystem requires experienced programmers.
4th Businesses are ready to invest in adapting their products and services with “AI-driven” solutions, even hiring people who can devise clever ways to reduce the cost of these API tokens.
Other than AI, for individuals, ways are under research on how to augment human capabilities with computers, and special-purpose robots for the elderly or disabled.
Automation in the workspace, e.g. all the support calls a company’s support gets converted to text, and the software does sentiment analysis on it, this can help categorize clients on an “emotion index” and help businesses understand what frustrates the client more.
Simple automation like advancements in vending machines like on-demand pizza or just roti, fresh tea, or coffee, is very common in Japan, and they need to continuously evolve to get a competitive edge. In countries like Pakistan where labor is cheap these solutions still do not make sense, but we have a simple vending machine at the office, and these will evolve with time.
AR/VR advancement will open a completely different realm of software that will be needed, which is specialized for their hardware. Consideration of spatial computing, apps that work well in 3D context and blend well with the physical world.
Legacy businesses that are still not completely digitized need custom software because they can not change how they do business. Every business will need an online equivalent, software is already available that helps B2B and B2C interactions, but people dont trust them as much. As more trusted companies are established in the market people will choose them over brick and mortar.
For programming language a new generation of language or tool is introduced to the market it makes development easier, and the barrier for entry is lowered, but that does never mean we don’t need experts in the previous tool.
e.g. when C language was developed it did not make assembly programming obsolete, it is still written to this date for low-level programming where resources are constrained (like IoT), in some cases, you can optimize c code with assembly but that is better done by the compiler. The same argument goes for Java it never made C or C++ obsolete. I dont have experience with low/no code tools, but similar efforts were made to generate code with UML models which did not stick. On this tangent, there were formal methods that aim generating code based on mathematical models which also failed due to its complexity.
Maintaining legacy systems, to start there are computer systems written on COBOL that still need support to this day, this is even more important for the internet-connected applications that need to be secure, for example, a web application running on Apache Tomcat will need to update itself for newer version if it wants to secure itself from known vulnerabilities.
To end, these advancements are cyclical, I remember that 20 years back in university someone claimed that all the software that the world needs is already developed, but still here in 2025 people are coding, debugging, and solving errors.
So if you like coding, explore areas in which you are interested, and develop expertise in it, I have not mentioned games, puzzles, and even 3D and 2D artwork can be automated by programming like generating a different color combination of the same artwork programmatically.
In Urdu we say “dana pani to likha waha hay” which somewhat translates as “food and necessities are written by the creator above” That is you will get your fair share if you keep working in whatever field you choose not just programming. A bit difficult to digest in this unfair world but we need more experts than cynics.