I think the big assumption that this article hinges on is that AI will never create durable code. That cannot be true. Now that we’ve finally got a taste of the singularity, humans will never stop trying to bring about AGI, which by definition must write code that is exponentially more durable than anything a human could write. The question is ‘when’.
In the 60s we thought we'd be living on the moon and mars "soon", flying cars were talked about seriously, that faster than sound jetliners were just around the corner, &c. Sometimes what's possible isn't desirable, or what's desirable isn't possible. No one went back on the moon, space programs are basically dead, planes are still as slow as they were in the 50s, a 747-8 from 2008 isn't much faster than a 707 from 1956
This library (including the schema documentation) was largely written with the help of Claude, the AI model by Anthropic. Claude's output was thoroughly reviewed by Cloudflare engineers with careful attention paid to security and compliance with standards. Many improvements were made on the initial output, mostly again by prompting Claude (and reviewing the results). Check out the commit history to see how Claude was prompted and what code it produced.
"NOOOOOOOO!!!! You can't just use an LLM to write an auth library!"
"haha gpus go brrr"
In all seriousness, two months ago (January 2025), I (@kentonv) would have agreed. I was an AI skeptic. I thought LLMs were glorified Markov chain generators that didn't actually understand code and couldn't produce anything novel. I started this project on a lark, fully expecting the AI to produce terrible code for me to laugh at. And then, uh... the code actually looked pretty good. Not perfect, but I just told the AI to fix things, and it did. I was shocked.
To emphasize, this is not "vibe coded". Every line was thoroughly reviewed and cross-referenced with relevant RFCs, by security experts with previous experience with those RFCs. I was trying to validate my skepticism. I ended up proving myself wrong.
Again, please check out the commit history -- especially early commits -- to understand how this went.
It took an expert and a whole team of software engineer to code "with IA" something well defined, well known, and with already lots of implementation in other languages
Glorious indeed, we have created a new kind of printer ! (irony inside)
(in conclusion, the above post from cloudflare is an ad)
Sorry, yes, talk of “AGI” or “the singularity” is silly and I should’ve distanced myself from that bit of the post. But the use of AI to write durable code is happening now, like it or not.
I think the big assumption that this article hinges on is that AI will never create durable code. That cannot be true. Now that we’ve finally got a taste of the singularity, humans will never stop trying to bring about AGI, which by definition must write code that is exponentially more durable than anything a human could write. The question is ‘when’.
> The question is ‘when’.
Why ?
In the 60s we thought we'd be living on the moon and mars "soon", flying cars were talked about seriously, that faster than sound jetliners were just around the corner, &c. Sometimes what's possible isn't desirable, or what's desirable isn't possible. No one went back on the moon, space programs are basically dead, planes are still as slow as they were in the 50s, a 747-8 from 2008 isn't much faster than a 707 from 1956
The question remains, 'can it be done'.
Just because you dream of something does not mean it will come true
Never stop trying != succeed
https://github.com/cloudflare/workers-oauth-provider
This library (including the schema documentation) was largely written with the help of Claude, the AI model by Anthropic. Claude's output was thoroughly reviewed by Cloudflare engineers with careful attention paid to security and compliance with standards. Many improvements were made on the initial output, mostly again by prompting Claude (and reviewing the results). Check out the commit history to see how Claude was prompted and what code it produced.
"NOOOOOOOO!!!! You can't just use an LLM to write an auth library!"
"haha gpus go brrr"
In all seriousness, two months ago (January 2025), I (@kentonv) would have agreed. I was an AI skeptic. I thought LLMs were glorified Markov chain generators that didn't actually understand code and couldn't produce anything novel. I started this project on a lark, fully expecting the AI to produce terrible code for me to laugh at. And then, uh... the code actually looked pretty good. Not perfect, but I just told the AI to fix things, and it did. I was shocked.
To emphasize, this is not "vibe coded". Every line was thoroughly reviewed and cross-referenced with relevant RFCs, by security experts with previous experience with those RFCs. I was trying to validate my skepticism. I ended up proving myself wrong.
Again, please check out the commit history -- especially early commits -- to understand how this went.
See the many comments on HN about this
It took an expert and a whole team of software engineer to code "with IA" something well defined, well known, and with already lots of implementation in other languages
Glorious indeed, we have created a new kind of printer ! (irony inside)
(in conclusion, the above post from cloudflare is an ad)
> TO EMPHASIZE THIS WAS NOT VIBE CODED
an expert using high powered autocomplete to successfully take a shortcut towards a well defined problem is not evidence of "AGI" or "the singularity"
belief in "the singularity" is a religious belief. you people sound like evangelicals talking about the goddamned rapture
Sorry, yes, talk of “AGI” or “the singularity” is silly and I should’ve distanced myself from that bit of the post. But the use of AI to write durable code is happening now, like it or not.
See also centuries of chrysopoeia: https://en.wikipedia.org/wiki/Chrysopoeia.
To be fair, they did synthesize gold in 1984, "although the production cost is estimated to be a trillion times the market price of gold."