🔥 Trending on HN

Why Some Open-Source Projects Say No to AI-Made Changes

3 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
open-source software(open-source software)

Software that people can use, inspect, and improve together.

pull request(pull request)

A proposed software change sent for review.

AI-generated contributions(AI-generated contributions)

Code, writing, or other project work made with AI tools.

What happened

Open Source Says No is a new directory about open-source software. It collects projects with public rules that restrict or reject AI-generated contributions. A November 11 snapshot was described as listing 22 projects. Examples included Immich, Meilisearch, Servo, QEMU, and Gentoo.

Each entry points readers toward a project’s contribution rules. It may show a short policy statement. The rules do not all say the same thing. Some ask contributors not to send AI-made pull requests. Others reject AI-generated code, tests, documentation, or issue reports. The directory therefore shows a range of policies, not one universal ban.

This story also became a topic on Hacker News. That attention shows community interest. It does not prove that the directory is complete or correct.

The background

Open-source projects accept proposed changes from outside contributors. A pull request is one common way to submit such a change. Maintainers then read it, test it, and decide whether it fits the project.

AI tools can produce code and explanations quickly. Human maintainers still need to check the result. They must understand what changed. They must check whether it fits existing design. They may also need to maintain it later. This creates a gap between the speed of making a contribution and the time needed to review it.

The policies collected by the directory reflect that problem. Some projects worry about code quality. Others worry about unclear authorship, maintenance, or the usefulness of AI-written reports. These concerns can lead to different rules for code, tests, documents, and discussions.

Why it matters

The important question is not simply whether AI is good or bad. The practical question is who carries the cost of checking AI-made work. A contributor may save time while a maintainer spends more time understanding the result. A project may decide that this exchange is unacceptable.

The directory can help contributors find rules before they submit work. It can also make differences between projects easier to see. However, it is only a signpost. The current policy on the project’s own contribution page remains the authority.

A project might reject outside AI-generated pull requests while allowing private AI assistance. Another might include documentation or issue reports in its restriction. Reading only the project name can hide these important differences.

What is confirmed

The directory confirms that multiple open-source projects have published restrictions on AI-assisted or AI-generated contributions. It also makes those policies easier to compare in one place. The November snapshot reported 22 entries, but that number belongs to a changing list.

The Hacker News discussion confirms that the topic attracted attention. It does not establish how many open-source projects share these rules. Hacker News points and comments measure attention, not the truth of the original policies.

What remains unknown

The list is not a census of open-source software. A project missing from it may welcome AI assistance, or it may simply have no published rule. A listed rule may also change before the directory updates.

The directory does not show whether bans improve code quality. It does not show whether they reduce maintainer workload. It also does not show whether contributors leave because of them. Those effects require separate evidence.

The meaning of AI-made work can also vary. One project may mean generated code. Another may include tests, documentation, images, or issue reports. The same label can hide different boundaries.

What to watch next

Contributors should use the directory as a starting point. They should then read the project’s latest contribution guide and submission template. They should check whether the rule covers code only, or all written contributions.

Maintainers may continue choosing among bans, disclosure rules, and human accountability. The next useful signal will be whether projects publish clearer policies and whether those policies reduce review noise. Open Source Says No is useful for tracking that change. It is not evidence that all open-source software says no to AI.

💬 OSS projects restricting AI contributions: a debate about review capacity and responsibility

Hacker News commenters do not reach a single verdict. Supporters frame restrictions as a way to protect limited review capacity and maintainers; critics say the relevant test is disclosure, understanding, testing, and responsibility—not whether an LLM was used. The site author describes the list as an early project built largely with AI, manually checked for correctness, and not intended to draw conclusions.

  • Supporters say the core problem is not generation itself but distinguishing a contributor who understands the domain and codebase from someone made overconfident by an LLM. They argue that cheap, low-effort PRs and issues can increase review load, cause design drift, and burn out maintainers; a commenter identifying as a Rust maintainer presents this as personal experience.
  • Critics argue that AI-assisted code is not automatically worse than human-written code. They favor disclosure and a contributor who reads, understands, tests, and takes responsibility for the submission. They also worry that blanket bans could exclude useful contributors from communities with limited funding and staff.
  • Several comments distinguish external AI-authored PRs from maintainers using AI themselves. A maintainer who knows the architecture can steer and verify the tool, so rejecting outside generated PRs may be a capacity policy rather than a claim that AI is always bad.
  • The labels may also be too coarse. One commenter says Immich rejects LLM-created PRs while its repository includes an agent file and Claude co-authored commits. On that account, a ban may mean no externally generated PRs, not no AI anywhere; each project’s policy needs to be checked individually.
  • A strong counterargument is security. Commenters say AI can help find bugs and vulnerabilities, while attackers may use it too. They therefore distinguish AI used for review, design, or security analysis from AI submitting code. The comments do not establish the effectiveness of either approach.
  • One user reports that a self-hosted tool that appeared to be AI-built left security issues reported on GitHub unfixed. That is an individual usage report, not a measurement of AI-built OSS as a whole.
  • The practical trade-off is among review time, project goals, contributor diversity, and security. The discussion mainly asks how to verify understanding and accountability, rather than proving that either blanket bans or blanket acceptance are universally correct.

initial digest at 75 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

🔥 Trending on HN

Some Open-Source Projects Reject AI-Made Changes

📰 Full story: Why Some Open-Source Projects Say No to AI-Made Changes

Some open-source teams reject changes made with AI tools. Here is why.

1 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
pull request(pull request)

A proposed change sent to a software project.

maintainer(maintainer)

A person who reviews and manages a software project.

Hacker News(Hacker News)

A website where people discuss technology news.

💡 The gist

  • A new website collects open-source rules about AI-made work.
  • The rules differ between projects and contribution types.
  • The list shows examples, not all of open source.

Open Source Says No is a website that collects these public rules. Open-source software is built so people can use and improve it. The site shows projects that restrict AI-made contributions.

A November 11 snapshot described 22 projects. Names included Immich, Meilisearch, Servo, and QEMU. Each entry points toward a project rule. Some rules ask people not to send AI-made pull requests. Others reject AI-made code, tests, documents, or issue reports.

A pull request is a proposed change. A contributor sends it to a project. Maintainers read the change before accepting it. They may test it too.

AI can create a change very quickly. A maintainer still needs time to understand it. The maintainer must check its behavior and its fit. This is the main reason behind many restrictions. Making work became faster. Checking work did not become equally fast.

The rules are not all-or-nothing. One project might restrict outside pull requests. Another might restrict every written contribution. A project may also ask for human responsibility instead of a total ban.

The story became a topic on Hacker News. Hacker News attention shows interest. It does not prove the list is complete or correct. Its points and comments measure attention, not accuracy.

The website is useful as a signpost. It is not an official rulebook for open source. A missing project may have no rule. Its policy may also have changed. Before sending work, read the project’s own contribution guide.

The key question is why the rule exists. It often concerns review time. AI may reduce writing time. People still need to check every change. That is why contributors should understand the rule before using an AI tool.

💬 Why some OSS projects reject AI-made contributions

The comments are divided. Some want to reduce a flood of hard-to-review changes; others say AI use should be acceptable when a human understands and owns the work. The page’s author says the list is an early project, manually checked, and not a conclusion.

  • Supporters say AI makes low-effort PRs cheap and harder to screen, which can waste review time and exhaust maintainers. A self-described Rust maintainer reports seeing this personally.
  • Critics say AI code is not automatically worse. They prefer disclosure, reading, understanding, testing, and responsibility; bans may push away useful contributors, especially from projects with limited resources, or make people hide their use.
  • Some projects may reject AI-generated PRs from outsiders while using AI internally, because maintainers know the codebase and can check the result. Commenters say the Immich example shows that a ban label may be too broad.
  • Others warn that a total ban could block AI-assisted bug and vulnerability finding. One user also reports a personal case of an AI-made-looking self-hosted tool with security reports left unfixed; neither point proves a general rule.
  • The debate is therefore about how to protect review capacity while keeping useful contributions and security work, not about a settled answer.

initial digest at 75 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

🔥 Trending on HN

Why Do Some Teams Say No to AI Changes?

📰 Full story: Why Some Open-Source Projects Say No to AI-Made Changes

Some computer-program teams do not want changes written by AI.

1 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
computer program(computer program)

Instructions that tell a computer what to do.

AI(AI)

A computer tool that can make things from instructions.

Hacker News(Hacker News)

A place where people talk about technology.

Open Source Says No is a website about shared computer programs. Many people can improve these programs.

Some teams do not want changes written by AI. AI can write changes quickly. People still need to read each change. Reading takes time.

Too many changes can tire a team. So some teams make a rule against AI-made changes. Other teams make smaller rules.

The website collects these rules. It does not make the rules itself. Each team chooses its own rule.

People talked about the website on Hacker News. Being popular does not make a rule true. The list can miss teams. Rules can change.

Read the team’s rule before sending a change.

💬 Can AI help change an open-source project?

People in the discussion disagree. The page maker says the list is small and new, and is not a final judgment.

  • Some project helpers say AI can send many changes that are quick to make but hard to check. Reading them can tire the people who care for the project; one maintainer describes this as a personal experience.
  • Other people say AI is okay when a human reads the work, understands it, tries it, and takes responsibility. A person-only program is not always better.
  • Some think AI can help find safety bugs, while one person tells of a tool that seemed AI-made and had reported safety problems that were not fixed. A project may use AI inside while refusing AI-made changes from outsiders. These are comments and one personal story, not proof that every AI project is good or bad.

initial digest at 75 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

Sources