It was the year 1878, and lots of people were talking about this new thing called electricity. Jude, owns a horse barn, he rents horses to other people, so he asks, “How can electricity help my business”, Watkins responds, “Electric lamps can help you light up your barn at night, it will make it easier for you to feed the horses at night”, and Jude responds, “But an oil lamp can already do that for me?”
Jude is right, electricity can help lighten up the horse ban, and could help with a few more routines in 1878, but you could also do most of those things without electricy. Of course, Jude could not have imagined, sitting in London and having a video call with someone in Australia, because electricity made computers and the internet possible.
At this point you are probably starting to get where I am getting to, yes, the immediate applications of electricity in 1878 could not have justified the scale of investment that eventually went into it, but so are the nature of GPTs( General Purpose Technologies), they transform the society in many interesting ways and affect every single aspect of our lives, but their impact is measured over time. In the case of electricity, over many years, electricity native applications such as the internet, advanced health devices used in medical centers all over the whole to diagnose health conditions, electric cars, computers etc have emerged as the true value of electricity. Today, no one questions anymore if building massive power stations are useful, we just know they are and we can’t live without them.
Today, we are back to the year 1878, a new GPT (General Purpose Technology) is here, it is called AI, championed by the rise of ChatGPT. Frontier LLMs such as GPT-5 have completely transformed the way we work and learn today. However, there are a lot of questions on whether the massive investments in the field can ever yield enough returns to justify the cost. A lot of this questions are related to lighting up the “Barn”, every application today is stacking an AI layer and its been great to see many useful integrations solving real pain points for individuals and business. However, we are just scratching the surface of the true value of AI.
Similar to electricity native technologies, AI native products are starting to emerge and will significantly transform the way we live, work and connect with one another.
A classical example of this is the evolution from AI completions, (e.g early github copilot, cursor tab) which basically made programming more enjoyable by providing better code autocomplete. Code completion is a classical example of using AI to improve an existing technology. However, the true value of AI for coding started to emerge with coding agents that can take a prompt as input and produce an entire app that actually works. Coding agents are the first great example of how much an AI native product can transform the society.
The first barely working coding agents arrived in 2024 with Cursor powered by Claude 3.5 making the waves. Today, GPT-5 with codex-cli has reached a level of capability most people would have predicted was like 5 years away. And this is just the beginning.
Frontier LLMs will ultimately enable anyone to bring any idea to life, not just in the domain of coding, but in core sciences and arts.
We are already seeing incredible progress with Maths. Advanced LLMs from OpenAI and Google Deepmind are already capable of solving maths and optimisation problems at expert human level.
This is the beginning of an era where we scientists can rapidly make new scientific discoveries in medical research, material science, energy production, agricultural science etc.
While we are not quite there yet, existing LLMs are already showing promise of being able to not just retrieve information that exists, but to generate novel solutions to previously unsolved problems, see [2509.18383] Gödel Test: Can Large Language Models Solve Easy Conjectures?
In addition to helping scientists solve problems that plague humanity today, access to quality education for every child in the world can become a reality as the cost of AI drops significantly, enabling low cost access to digital agents that can teach at human expert level, but in a way that is personalised to every child in the world.
Growing up in rural Nigeria, I had to attend schools where the quality of eduction was significantly low, I relied a lot on self learning, Wikipedia and google search to help me gain the knowledge I needed to thrive. I can only imagine what access to even the free ChatGPT would have done for me as a teenager.
To make this possible, an energy and compute infrastructure build out like the world has never seen before is necessary to bring the true promise of AI to the over 7Billion people on planet earth. A key reason is, LLMs leverage computation at scale to solve problems, and as the scaling laws have shown, if we can put in more and more compute into LLMs, we can solve more and more problems, and the ceiling on this, remains to be seen.
Yes, today, AI bolted onto a wide variety of business and consumer applications may not justify the 500 Billion dollars Stargate project OpenAI is doing at the moment, but the value that will be created by AI native solutions will dwarf that cost.
What this means for companies, startups and everyone is, we need to think and understand the change around us in the context of the internet rather than electric lamps. The world we know today, will be completely transformed in a few years. The nature of companies and jobs will change, and I believe more good will be done, more opportunities created than ever before. It is true there are challenges, especially with the nature of jobs. Already job losses are a big concern as AI coding agents becomes much more capable than junior engineers. Each engineer can create 10x more than they could before, and in some ways, that can be translated to greater abundance. But it also means less engineers per task.
One reason to be excited despite the challenges is the rise again of the golden age of science. As of a few years ago, everyone did software whether they loved it or not, because there was not much money in being a physicist or chemist or biologist. Fundamental sciences has received far less attention that it needed. With AI automating software and simultaneously making it easier to scale up science research, there is a lot more work and hopefully jobs in finding answers to fundamental questions about the nature of the universe, the human body, and the natural laws of physics that govern our everyday life
Summary
AI is changing the world, and there is no going back to an age where machines could not speak back to us or perform tasks like an intelligent human, we need enormous energy and compute to unlock the value of this amazing technology, its not all going to be smooth ride, but as a society, we can make it work for us just as we made electricity work for our betterment.
Onto AGI and Superintelligence!