Piousbox

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Okay... not to diss on the author, but this hasn't aged well over the last 3 years. As a prorammer, I use AI every day. I used 3 of them in the past hour. Sure, I check every line AI generates, but I use AI nonetheless. Sure, I often discard the AI-generated changes and start over (and over, and over), yet it's still faster than me trying to type it all by hand. I'll literally tell AI to try to do something, scrap it, and try again, half a dozen times in a row, with slight modifications, and that will be faster than doing it myself.

Not to mention that oftentimes the AI does work! See this quick article for an example: https://wasyaco.com/w/20260815

Would I have come up with this solution on my own? I doubt it. Does it save me time and improve efficiency? Absolutely.

AI is very powerful, and works better than ever, and continues improving. Sure, a lot of it is fraud, a lot of it is questionable, and even the best modern AI will try to ruin your codebase, if you let it. (A junior programmer would also ruin your codebase if you let him.)

Fundamentally, I believe the GPT architecture actually go it right. The way brain works, the way neurons work. Each neuron does integrals, and together they *predict*, anticipate something, and that anticipation is what builds the world around us. The "attention" mechanic that divides computation into key, query, value, seems to work so well. And arguably, there is no cognition outside language. GPT does very well with language, with plain text, and generating (and consuming) plain-text instructions. Amazingly, even if all GPT does is give you probability of the next token, this alone is so powerful, that it looks like cognition.

So yeah. We are at the very scary place where AI *is* becoming indispensable. Even in defending yourself against AI.

Well, is it?