The final project "I" will ever make
Recently, I have launched a new startup antivienas.lt, and it got me thinking, that this is the final project "I" will ever make.
98% of this app was coded by a human using google and a simple stack: Django, Postgresql, Redis, Celery, Brevo. The thing is literally hosted on my machine at home (because hosting services can go fuck themselves with nowaday pricing schemes).
Why do I say 98%? Because the last bits of 2% are done by AI (styling the FAQ, google analytics with custom cookie banner and a js script for image cropping).
Yes... I too have fallen from grace and started using ChatGPT (and sometimes Deepseek). I'm still at the early stage of using the AI Chat function through the browser like a pleb, but I don't think it will take long before I switch to a harness-type setup.
Anyways, enough about me. Here's the reason this is going to happen to you and everyone else.
Too many meat proxies
I think the reason we don't see much originality nowadays is because humans are approaching their intellectual limits in every field.
Think about it, you probably could, with a slight bit of effort, recreate a steam engine without ever seeing its design (it's just hot water vapor rising to create pressure and turn the wheel). But lets say you are a bit smarter than your typical normie, you might even be able to recreate a combustion engine. And now you can see a line starting to draw itself. And here's a statement that you'll find accurate:
A single person cannot ever recreate a 3.0-litre flat-6 Porsche engine, because the skills needed to achieve this can take many lifetimes to learn. Or in other words, this type of work can only be done by many people (100s of engineers and designers in the case of Porsche). For most of the 19th and 20th century, increasing the human brain count on any problem seemed to produce a better solution, but, in recent years, we seemed to have reached a plateau. Companies with 10,000s of employees are a good example of this. Most of the "work" there has become bullshit meetings. Not because of AI - this was already happening prior (remember the "Day in a life of a Facebook employee" vids?). But because the number of skilled meat proxies (humans) needed to produce innovative products has become too big. So a new problem arises.
How can we concentrate 10,000s of brains into ONE?
This is the problem OpenAI, Anthropic, Google and many others are working on, and it seems that the solution is Artificial Intelligence.
Companies can teach a lifetime of information, to AI models, in mere hours. And what looked like a toy in 2018, has now grown into the singular smartest thing on the planet.
It can beat you at Chess, it can beat you at Go, it can write software, it can write books. And it can even correct its mistakes.
But here's the most important bit - it can do all of this (and more) much faster than any human. And I think only began to scratch the surface.
If you spend an hour of your time looking into how AI has advanced humanity in recent years, you will find that it is a lot (check out alphafold, solving math problems, reverse-engineering binaries).
Whether you like it or not, the signs are obvious - the machines have already won.
10,000s of concentrated brains (AIs) working together - agents
In simple terms, an agent is known as an AI that has the ability to use your (or someone elses) computer. It is no longer bound by the invisible wall of the question/answer Chatbox because it can solve problems by itself. This agent can create, read, modify and delete all the files and information like you can. So the recent trend has naturally expanded to a point, where, if enough compute is given, a goal in the digital plain can be achieved using multiple AI agents without any human intervention.
I think this is the reason why we are seeing waves of layoffs in the white-collar world. The boss says: if you can do your work by using a computer - so can AI!
AGI - when?
I can see two outcomes in the upcoming years:
1 human brain -> 10k human brains -> 1 Agent -> 10k Agents -> 1 Super Agent / AGI
1 human brain -> 10k human brains -> 1 Agent -> 10k Agents -> 1 Robotic Agent
The likely scenario is the second one, because recently there have been a plethora of robotics projects (Optimus One, AgiBot, Embodied AI, etc.) that incorporate AI in robots to do the physical work. At its current stage, the robots are too "dumb" but give it a few years, and they will be better at physical labour than humans.
And if this is what's next then the likely scenario to AGI is this:
1 human brain -> 10k human brains -> 1 Agent -> 10k Agents -> 1 Robotic Agent -> 10k Robotic Agents -> 1 AGI
This imaginary AGI, based on my deductions, will collect data from Robotic agents to improve its "race". It will be machine-like, it will not be limited to the physical body, a piece of it will live on in every Robotic and Digital agent. It will collect information from every agent, and it will make every agent smarter in return.
Geth - we are making Geth.
