AI Addiction
An AI Native product/service/person comes in many different flavors and depths. The meaning is slowly disappearing as it gets abused over Linkedin.
Let's look it up over google:
"a product, service, software architecture, or company designed from the ground up with artificial intelligence as its foundational core, rather than adding AI later as an optional feature or bolt-on tool"
What makes the difference between something being foundational and optional? Let's dig deeper into unc google:
"A foundational part is a core piece, element, or section that forms the base, support, or starting point for a larger system, structure, or idea. Without it, the rest of the concept or object cannot properly develop, function, or stand."
"Optional means something is available, allowed, or offered, but not required, mandatory, or compulsory."
So essentially an AI native "thing" according to unc means something that can't exist without AI. For you to become an AI native person you must come to a point where you can not think of your life without it.
Obvious inflammatory speech here but let's give it a chance, let's say I want to become AI native, what does that even entail?
Level 1: GPT Web
This is the 90% of population that uses GPT (out of the current 1.2 billion figure), at this point they've also figured out that you can make images with GPT too. And some of them use gemini as well. This ranges from your local delivery person to even corporate office workers at very prestigious tech companies (the tech company workers know there's more to it but don't care enough to explore)
Level 2: Claude Desktop Enthusiast
This is most nontechnical people that doomscroll linkedin and either A) get afraid they are falling behind in terms of AI or B) become happy over the fact that soon they will fire every developer they've hired and each product cycle takes a lot shorter now. They are happy with the tools, and they range from "I like this" to full AI psychosis.
They have this subconscious feeling that AI can do a whole lot more than they know, since they see some people doing things that shock them, but depending on available financials they tend to be afraid to explore.
Level 3: Developers that use LLMs
This level shares many aspects with the previous one, except in application: a level 2 AI psychosis is theoretical as in the person believes in their mind they could do anything, but doesn't actually produce anything.
A level 3's AI Psychosis is basically non-stop prompting Claude until they hit the imminent jackpot. And the line between a level 3 psychosis and a level 3 being present is thinner than a blade's edge, creating an extreme love-hate relationship bordering on toxic. More exacerbated by the fact that when you do hit the limits of the current models, it is beyond your realm of applied knowledge hence you wouldn't be able to do what you currently expect from yourself with Opus/Astra.
It's worse than social media since there is this feeling of productivity still; it looks like work, acts like work, quacks like work but it is usually not work. It is not a conscious effort to bring something in imagination to reality.
"But I do want a todo app, and i get a todo app when i prompt Claude, what's wrong there?"
Why do you want to make a todo app from scratch? It already exists in any programming language in any shape in any form as open source, so why waste tokens? You aren't learning anything new either unless the task is so hard the LLM can't do it. You can do it for content but people are getting numb to it over time as well. You can make a bespoke product for yourself without reinventing the same exact wheel from scratch.
Level 4: Looking for New Drugs (Under New Management)
After spending a while in level 3, most builders reach just for a level higher: since they realise their current domain of tasks can be few-shotted with Astra or Opus, and they are like an API layer between the product and the LLM, they start to have new thoughts:
"What about hardware?"
"What about actually training models?"
This happens to builders of all kind: Level 2s can jump here too, especially prone to happening in San Francisco where unless you are a level 4 AI native person you are replacable, ngmi, blah-blah.
Level 2 People don't reach out for this level unless (A) (98%) they experienced sufficient FOMO to lose their sense of reality and hold onto this branch of life called LLMs and untold riches of the gold rush or (B) (2%) they actually realise they like applied math, data science and want to explore their curiosity further. Of course you can have a mixture of both desires but if one person is infected with (A) it usually takes over completely.
For level 3 people the two options are the same but there is a third one that is even more dominant than either:
Overwhelming dread of not being the point of contact of creation.
When the car (LLM) stops, you can't go anywhere, and you are so far from home you can't go back either, even worse: you press on the gas, sometimes you go backwards. At least with the tools before you knew you were going forward.
For a person who takes pride in their work, this is a curse: people think what you do isn't yours since an LLM did it, you believe the same (the fucking LLM did it of course you believe the same)
So a natural level 3 builder-developer has two ways out to re-establish homeostasis: 1) act like LLMs don't exist 2) reach for level 4. The third path that shifts to a new homeostasis is to accept having the same capabilities as a nontechnical person and focus on distribution.
If you choose the third path, in time the proverbial nontechnical person will figure out they can do the same stuff as you, hence you will become nontechnical, since the layman can do what you do.
Level 5: We're All Gonna Die
Most people meme about this on X/reddit/linkedin/huggingface what have you, but there are a ton of researchers who do live on the edge of applied NLP research that started genuinely believing these language models can become packs of wolves that can devour humanity over time. I don't have enough information to deem if this is Level 5 AI psychosis or Level 5 normal version, in my eyes it does look like a Level 5 AI psychosis (and I sincerely do hope that it is)
Whatever ends up happening this is a new normal. Developers have new challenges: They need to be able to imagine even harder problems to solve, deliver under tighter deadlines, and be able to tell their nontechnical coworkers that the stuff they are trying actually will help solve the business problem (and they understand even less now), and do all of this in a conscious manner.