I had PR #4093 rejected as "AI generated" and I think this decision misunderstands what actually happened. I'm raising this publicly because I want clarity on the policy and I believe this rejection was incorrect.
I've been working on a list of historical map resources since at least January 2023, and one of my main sources was maphistory.info, a site run by Tony Campbell (former Map Librarian at the British Library) that's been the authoritative gateway for history of cartography since 1997. I have had a lot of life events what left me with little bandwith to figure out the linter (I am a humanities person, all tech work is self-taught), so I didn't submit for 3 years.
On 1 January 2026, Tony Campbell announced the site was shutting down after 30 years. So in Feb, I used Claude Code to overcome the linting barrier so I could finally submit some preservation work before the resource disappeared completely as well. This included setting up the GitHub page and reformatting it for the linter as I had a number of design choices that did not pass. It tagged those commits with "Co-Authored-By: Claude" to be transparent about using AI for the technical formatting. You can see this here in my first commit of 2026: https://github.com/stark1tty/awesome-historical-maps/commit/2acc2f0f334a40cd920fcc58be8a8e979181232e
For the most recent batch of links, I went through pages from maphistory.info, checked which links were still alive, and selected the ones worth preserving. I also dug out other links I have been passively collecting. If that's not manual curation, I don't know what is. It's literally curated content derived from a curated directory and I can provide the spreadsheets I am working from. There are thousands of links in my files I have not gotten to yet for obvious reasons.
My PR was rejected with "AI generated lists are not accepted" and "All your replies here are clearly AI generated too."
The commit history shows manual work: Feb 2023-Jan 2026 commits are incremental manual additions. The rep shows nearly 200 commits, most small and incremental, i.e. several one line changes within an hour.
I used Copilot to help walk me through this pull and grab my history, I can provide the conversation as well, but it is just me trying to get the linter to work correctly and pass this off my plate. AI was used to format markdown and set up Jekyll so I could meet linting requirements, not to generate or find the curated content.
I quite literally research this topic academically and professionally and have been contributing to other Awesome lists incrementally for years.
The list is blatantly asymmetrical in coverage and specificity, AI does not output that kind of work.
So, I am asking:
Is this rejection correct under the policy? If AI assistance for overcoming technical barriers disqualifies 3 years of manual curation and link verification, that needs to be explicit. The linter forced me to remove my human touches, so I'm not sure what you want. I would be more than happy to add some back.
What evidence is sufficient? I provided verifiable sources, commit history, archive snapshots showing pre-AI existence, and offered source files. What else could possibly demonstrate manual curation?
Did transparency backfire? Would hiding the AI assistance have resulted in acceptance?
This is partially a preservation work of a significant scholarly resource that just shut down and I am also doing it for accessibility of these resources. I think this deserves reconsideration, but at minimum I need clarity on what the actual rules are.
In addition, issue 732 expresses a need for non-tech lists, yet this rejection suggests that if a Humanities expert uses developer tools to meet technical standards, their domain knowledge is dismissed as 'AI-generated.' This creates a barrier for the very contributors that issue claims to welcome.
Originally created by @stark1tty on GitHub (Apr 12, 2026).
Original GitHub issue: https://github.com/sindresorhus/awesome/issues/4094
I had PR #4093 rejected as "AI generated" and I think this decision misunderstands what actually happened. I'm raising this publicly because I want clarity on the policy and I believe this rejection was incorrect.
I've been working on a list of historical map resources since at least January 2023, and one of my main sources was maphistory.info, a site run by Tony Campbell (former Map Librarian at the British Library) that's been **the** authoritative gateway for history of cartography since 1997. I have had a lot of life events what left me with little bandwith to figure out the linter (I am a humanities person, all tech work is self-taught), so I didn't submit for 3 years.
On 1 January 2026, Tony Campbell announced the site was shutting down after 30 years. So in Feb, I used Claude Code to overcome the linting barrier so I could finally submit some preservation work before the resource disappeared completely as well. This included setting up the GitHub page and reformatting it for the linter as I had a number of design choices that did not pass. It tagged those commits with "Co-Authored-By: Claude" to be transparent about using AI for the technical formatting. You can see this here in my first commit of 2026: https://github.com/stark1tty/awesome-historical-maps/commit/2acc2f0f334a40cd920fcc58be8a8e979181232e
For the most recent batch of links, I went through pages from maphistory.info, checked which links were still alive, and selected the ones worth preserving. I also dug out other links I have been passively collecting. If that's not manual curation, I don't know what is. It's literally curated content derived from a curated directory and I can provide the spreadsheets I am working from. There are thousands of links in my files I have not gotten to yet for obvious reasons.
My PR was rejected with "AI generated lists are not accepted" and "All your replies here are clearly AI generated too."
Why I think this is wrong:
- This is a 3-year old project that existed before Claude Code existed: [[Internet Archive snapshots from 2023-2025](https://web.archive.org/web/*/github.com/stark1tty/awesome-historical-maps)](https://web.archive.org/web/*/github.com/stark1tty/awesome-historical-maps) show the repo and content predating the AI assistance. **The base of this project was set up before GPT had web search access.**
- The commit history shows manual work: Feb 2023-Jan 2026 commits are incremental manual additions. The rep shows nearly 200 commits, most small and incremental, i.e. several one line changes within an hour.
- The source is real and verifiable: The maphistory.info shutdown announcement is [[publicly documented](https://web.archive.org/web/20260101000000*/maphistory.info)](https://web.archive.org/web/20260101000000*/maphistory.info) and you can compare the links.
- I used Copilot to help walk me through this pull and grab my history, I can provide the conversation as well, but it is just me trying to get the linter to work correctly and pass this off my plate. AI was used to format markdown and set up Jekyll so I could meet linting requirements, not to generate or find the curated content.
- The intent to do these things to this repo was listed in early 2023: https://github.com/stark1tty/awesome-historical-maps/commit/481ca0f78d0f7b3180e0f7fa04a4ba930c1f4332
- I quite literally research this topic academically and professionally and have been contributing to other Awesome lists incrementally for years.
- The list is blatantly asymmetrical in coverage and specificity, AI does not output that kind of work.
So, I am asking:
1. **Is this rejection correct under the policy?** If AI assistance for overcoming technical barriers disqualifies 3 years of manual curation and link verification, that needs to be explicit. The linter forced me to remove my human touches, so I'm not sure what you want. I would be more than happy to add some back.
2. **What evidence is sufficient?** I provided verifiable sources, commit history, archive snapshots showing pre-AI existence, and offered source files. What else could possibly demonstrate manual curation?
3. **Did transparency backfire?** Would hiding the AI assistance have resulted in acceptance?
This is partially a preservation work of a significant scholarly resource that just shut down and I am also doing it for accessibility of these resources. I think this deserves reconsideration, but at minimum I need clarity on what the actual rules are.
In addition, issue 732 expresses a need for non-tech lists, yet this rejection suggests that if a Humanities expert uses developer tools to meet technical standards, their domain knowledge is dismissed as 'AI-generated.' This creates a barrier for the very contributors that issue claims to welcome.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @stark1tty on GitHub (Apr 12, 2026).
Original GitHub issue: https://github.com/sindresorhus/awesome/issues/4094
I had PR #4093 rejected as "AI generated" and I think this decision misunderstands what actually happened. I'm raising this publicly because I want clarity on the policy and I believe this rejection was incorrect.
I've been working on a list of historical map resources since at least January 2023, and one of my main sources was maphistory.info, a site run by Tony Campbell (former Map Librarian at the British Library) that's been the authoritative gateway for history of cartography since 1997. I have had a lot of life events what left me with little bandwith to figure out the linter (I am a humanities person, all tech work is self-taught), so I didn't submit for 3 years.
On 1 January 2026, Tony Campbell announced the site was shutting down after 30 years. So in Feb, I used Claude Code to overcome the linting barrier so I could finally submit some preservation work before the resource disappeared completely as well. This included setting up the GitHub page and reformatting it for the linter as I had a number of design choices that did not pass. It tagged those commits with "Co-Authored-By: Claude" to be transparent about using AI for the technical formatting. You can see this here in my first commit of 2026: https://github.com/stark1tty/awesome-historical-maps/commit/2acc2f0f334a40cd920fcc58be8a8e979181232e
For the most recent batch of links, I went through pages from maphistory.info, checked which links were still alive, and selected the ones worth preserving. I also dug out other links I have been passively collecting. If that's not manual curation, I don't know what is. It's literally curated content derived from a curated directory and I can provide the spreadsheets I am working from. There are thousands of links in my files I have not gotten to yet for obvious reasons.
My PR was rejected with "AI generated lists are not accepted" and "All your replies here are clearly AI generated too."
Why I think this is wrong:
So, I am asking:
Is this rejection correct under the policy? If AI assistance for overcoming technical barriers disqualifies 3 years of manual curation and link verification, that needs to be explicit. The linter forced me to remove my human touches, so I'm not sure what you want. I would be more than happy to add some back.
What evidence is sufficient? I provided verifiable sources, commit history, archive snapshots showing pre-AI existence, and offered source files. What else could possibly demonstrate manual curation?
Did transparency backfire? Would hiding the AI assistance have resulted in acceptance?
This is partially a preservation work of a significant scholarly resource that just shut down and I am also doing it for accessibility of these resources. I think this deserves reconsideration, but at minimum I need clarity on what the actual rules are.
In addition, issue 732 expresses a need for non-tech lists, yet this rejection suggests that if a Humanities expert uses developer tools to meet technical standards, their domain knowledge is dismissed as 'AI-generated.' This creates a barrier for the very contributors that issue claims to welcome.