CHAOSS OSPO WG, March 19 2026 - CourtneyR.dev's 2nd Brain

CHAOSS OSPO WG, March 19 2026 ![rw-book-cover](https://img.youtube.com/vi/H9SsKPdj80Y/maxresdefault.jpg) 🔗 **Source:** [CHAOSS OSPO WG, March 19 2026](https://www.youtube.com/watch?v=H9SsKPdj80Y) > [!note]+ Document Notes > tags: #open-source #AI #OSPO #community-health #dependency-management > author: CHAOSS OSPO WG > source: YouTube > reference: https://www.youtube.com/watch?v=H9SsKPdj80Y > date created: 2025-01-11 10:30 > > # 🏷️ Title: AI and Open Source Alignment: What OSPOs Should Be Thinking About >  ## 🧰 Action Inbox - [H1] Join CHAOSS AI alignment working group for ongoing research and resources — 45:23 #AI-alignment #community - [H2] Develop repository contribution guidelines to manage AI-generated PR volume — 35:47 #repository-management #automation - [H3] Create context standards for agents to ensure open source compliance — 15:32 #governance #licensing ## 🧭 Table of Contents - [[#📝 Summary]] - [[#📖 Main Content]] - [[#🗒️ Notes]] - [[#📚 References]] - [[#🗣️ Paraphrase]] - [[#📰 Newsletter]] - [[#✅ Action Items]] - [[#🧪 Hypotheses]] - [[#📊 Methodology]] - [[#❓ Questions]] - [[#🔮 Future Research]] - [[#💡 Implications]] - [[#🧠 Learning Styles Assessment]] - [[#📊 Accessibility Assessment]] - [[#🌍 DEIB Assessment]] - [[#✅ Logical Validity]] - [[#🛡️ DARVO Analysis]] - [[#🎯 Conclusion]] ### 🔗 References - [CHAOSS AI Alignment Working Group](https://github.com/chaoss/wg-ai-alignment) - [OSI Open Source AI Definition](https://opensource.org/blog/the-open-source-ai-definition-is-here) ## 📝 Summary - **What happened**: Emma Irwin presented on AI alignment challenges for OSPOs, covering use, release, and contribution workflows - **Why it matters**: AI is fundamentally changing how open source is built, released, and sustained, requiring new OSPO strategies - **Key stakeholders**: OSPO teams, maintainers, engineers using AI tools, open source communities - **Risks and tradeoffs**: Increased maintainer burden vs. productivity gains; licensing compliance vs. speed; human oversight vs. automation - **Bottom line**: OSPOs must ensure AI systems remain aligned with human values and community expectations while managing new risks ## 📖 Main Content Emma Irwin, formerly of Mozilla and Microsoft OSPOs, presented findings from the CHAOSS AI Alignment Working Group on how AI is transforming open source workflows. AI alignment, also known as AI safety, focuses on ensuring AI systems do what humans want and need rather than operating independently of human values. The presentation covered three key areas where AI intersects with traditional OSPO responsibilities: **Using Open Source with AI**: Dependencies are increasingly selected by AI agents, creating new risks around typo-squatting, licensing compliance, and security. Traditional metrics for community health may become misleading as AI activity masks human contributor burnout. **Releasing Open Source**: Organizations face challenges around code written by AI agents, unclear licensing provenance, and the need for updated governance policies. The definition of "open source AI" remains contested, with various interpretations from truly open models to vendor marketing claims. **Contributing to Open Source**: Maintainers report increased burden from AI-generated pull requests, often poorly targeted or formatted. Some projects are implementing stricter contribution guidelines to manage the volume while preserving quality. The working group emphasizes that alignment isn't just technical but social - ensuring AI serves both community values built over decades and legitimate business interests in open source. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/H9SsKPdj80Y?t=0) — [H1] Introduction and Background - Key idea: Emma Irwin introduces 20 years open source experience, 10 as developer, 10 as technical program manager - Evidence: "I've been working in open source for 20 years... helping organizations be efficient and respectful and show up in the open source ecosystem" #### ⏱️ [02:15](https://youtu.be/H9SsKPdj80Y?t=135) — [H2] AI Alignment Definition - Key idea: AI alignment ensures systems do what humans want and need, not just what they're programmed to do - Evidence: "Alignment is really about ensuring that what AI systems do is what humans want and need essentially" #### ⏱️ [04:30](https://youtu.be/H9SsKPdj80Y?t=270) — [H1] Two Contracts in Open Source - Key idea: Open source has always had separate value propositions for businesses and communities that may be diverging - Evidence: "We've always had those kind of two contracts essentially... my worry is it's getting a little bit too further apart the more we remove humans" #### ⏱️ [07:45](https://youtu.be/H9SsKPdj80Y?t=465) — [H2] Using Open Source - New Risks - Key idea: AI introduces novel dependency risks including hallucinated package names and unclear licensing - Evidence: "There's you know just kind of a leveling up of how there's both the hallucination of names now coming into play" #### ⏱️ [10:20](https://youtu.be/H9SsKPdj80Y?t=620) — [H3] Code Commits by AI - Key idea: Significant portion of code commits are now authored by AI agents alone - Evidence: "There's 4% of Claude commits are now authored by Claude alone" #### ⏱️ [12:15](https://youtu.be/H9SsKPdj80Y?t=735) — [H1] Monitoring Community Health - Key idea: Traditional metrics may mislead as AI activity masks human contributor burnout - Evidence: "We actually might see an increase in merges... which makes it seem like there's more contributors when it's really AI" #### ⏱️ [15:32](https://youtu.be/H9SsKPdj80Y?t=932) — [H2] Shaping Agent Behavior - Key idea: OSPOs need to provide context to AI agents to avoid biased recommendations toward proprietary solutions - Evidence: "If I ask an agent to build an open source solution it'll immediately go for llama... which actually are mixed license" #### ⏱️ [18:45](https://youtu.be/H9SsKPdj80Y?t=1125) — [H3] Governance and Consent - Key idea: Code of conduct updates needed to address AI use and community consent - Evidence: "We're adding an AI clause around consent... anytime one of our maintainers uses AI is with the consent of the community" #### ⏱️ [21:30](https://youtu.be/H9SsKPdj80Y?t=1290) — [H2] Abandoned Dependencies Question - Key idea: Debate over whether AI agents should maintain abandoned open source projects - Evidence: "Should agents manage abandoned dependencies?... is that something we want to happen? What does that look like?" #### ⏱️ [25:15](https://youtu.be/H9SsKPdj80Y?t=1515) — [H3] Repository Hygiene for Agents - Key idea: Repositories need documentation to guide AI development in alignment with project standards - Evidence: "Recently some of our repos shipped like agents.md to be able to facilitate work with agents doing development" #### ⏱️ [28:40](https://youtu.be/H9SsKPdj80Y?t=1720) — [H1] Open Source AI Definition Confusion - Key idea: Market confusion around what constitutes "open source AI" with various interpretations - Evidence: "There's the OSI definition... there's the meta llama style there's vendor marketing... when I ask what open-source AI solutions do you have I have to really keep reprompting" #### ⏱️ [32:20](https://youtu.be/H9SsKPdj80Y?t=1940) — [H2] Copyright and Licensing Challenges - Key idea: AI-generated content is generally non-copyrightable, creating licensing uncertainty - Evidence: "Scholars and practitioners are converging towards the fact that AI generated stuff is non-copyrightable" #### ⏱️ [35:47](https://youtu.be/H9SsKPdj80Y?t=2147) — [H1] Managing PR Volume - Key idea: Projects implementing stricter guidelines to manage AI-generated pull request burden - Evidence: "We don't accept PRs that don't... address an existing issue, touch more than a certain amount of files... are untested" #### ⏱️ [40:15](https://youtu.be/H9SsKPdj80Y?t=2415) — [H2] Maintainer Crisis Worsening - Key idea: AI is exacerbating existing open source sustainability problems - Evidence: "The maintainer crisis is getting worse... it's causing a lot more problems" #### ⏱️ [43:30](https://youtu.be/H9SsKPdj80Y?t=2610) — [H3] Decreased Funding Trends - Key idea: Fewer companies contributing financially to open source despite increased need - Evidence: "There's fewer and fewer companies contributing to open source financially... it's decreased, it's just going down" #### ⏱️ [45:23](https://youtu.be/H9SsKPdj80Y?t=2723) — [H1] Call to Action - Key idea: Join CHAOSS AI alignment working group for collaborative learning and resource sharing - Evidence: "Join our AI alignment working group chat... there's lots of great links being shared" ## 📚 References - [CHAOSS AI Alignment Working Group](https://github.com/chaoss/wg-ai-alignment) - [OSI Open Source AI Definition](https://opensource.org/deepdive-drafting/) - [Fair Source Initiative](https://fair.io/) - [Contributors Covenant](https://www.contributor-covenant.org/) ## 🗣️ Paraphrase - AI is fundamentally changing how we use, release, and contribute to open source software, requiring OSPOs to develop new strategies for governance, compliance, and community health - Traditional metrics for measuring open source community health may become misleading as AI activity can mask human contributor burnout while appearing to show increased engagement - OSPOs need to proactively set context and guidelines for AI agents to ensure they recommend truly open solutions rather than falling back on biased training data that favors proprietary options ## 📰 Newsletter **AI Alignment Challenges for Open Source Program Offices** The integration of AI into open source workflows is creating new challenges that require immediate OSPO attention. From dependency selection by AI agents to managing the flood of AI-generated pull requests, traditional open source governance models need updating. Key areas requiring attention include establishing context standards for AI agents, updating contribution guidelines to manage PR volume, and developing new metrics that can distinguish human from AI activity. The CHAOSS AI Alignment Working Group is actively researching these challenges and building resources for the community. **Resources**: [CHAOSS AI Alignment WG](https://github.com/chaoss/wg-ai-alignment), [Learn WordPress](https://learn.wordpress.org/), [Developer Docs](https://developer.wordpress.org/), [WP-CLI](https://wp-cli.org/) ## ✅ Action Items #### DevRel - Developer advocacy: Create educational content on AI tool usage in open source contexts, metrics for success include workshop attendance and resource downloads - Education: Develop training materials on AI alignment principles for open source developers, establish office hours for AI-related questions - Developer experience: Update documentation to include guidance for AI agents working with repositories - Community management: Moderate discussions around AI tool usage policies, gather feedback on community preferences #### OSPO - Policy & Governance: Draft AI usage policies for code contributions, establish consent frameworks for AI tool usage in projects - License Compliance: Develop tooling to track AI-generated code provenance, create compliance review processes for AI-authored contributions - Security Management: Assess risks from AI-selected dependencies, implement monitoring for AI-introduced vulnerabilities - Software Inventory: Enhance SBOM tracking to include AI-generated components, establish ownership clarity for AI-authored code - Developer Support: Create training on responsible AI tool usage, provide consultation on AI governance policies - Community Engagement: Participate in CHAOSS AI alignment working group, collaborate on industry standards for AI in open source #### Community - Maintainers: Develop contribution guidelines for AI-generated PRs, create roadmap alignment strategies to prevent drift - Contributors: Establish consent protocols for AI tool usage, provide education on responsible AI contribution practices - Sponsors: Fund maintainer support specifically for AI-related workload increases, sponsor research into community health metrics #### WordPress - Core: Evaluate AI tool impact on contribution quality, establish policies for AI-generated patches - Core Performance: Monitor AI-generated code for performance impacts, develop benchmarks for AI-authored optimizations - Documentation: Create guidance for AI tool usage in WordPress development, update contributor handbooks - Community: Survey community attitudes toward AI tools, develop consensus on acceptable AI usage - Training: Integrate AI tool best practices into existing curricula, create specific courses on responsible AI usage - Support: Develop escalation paths for AI-related issues, train support staff on AI tool limitations ## 🧪 Hypotheses - AI agent selection of dependencies will increase security risks due to bias toward popular but potentially vulnerable packages - Traditional community health metrics will become unreliable as AI activity masks human contributor burnout and disengagement - Projects with explicit AI usage policies and contribution guidelines will maintain higher code quality than those without such governance - OSPOs that proactively address AI alignment will see better long-term community relationships and contributor satisfaction ### 📊 Methodology - Survey OSPO practitioners on current AI tool policies and observed impacts on community health metrics - Analyze commit patterns and PR characteristics to distinguish AI-generated from human-authored contributions across sample projects - Compare community satisfaction and maintainer burnout rates between projects with and without AI governance policies ### ❓ Questions - How can OSPOs effectively measure the true health of human contributors when AI activity dominates traditional metrics? - What governance frameworks best balance AI productivity benefits with community consent and maintainer autonomy? - How should organizations handle liability and accountability when AI agents make decisions about security-critical dependencies? ### 🔮 Future Research - Development of AI-aware community health metrics that can distinguish human from automated contributions - Investigation of optimal governance models for AI agent behavior in open source contexts - Long-term studies on the impact of AI tools on maintainer burnout and community sustainability ### 💡 Implications - Practice: OSPOs must immediately begin developing AI governance policies and updating existing processes for the AI era - Policy: New frameworks needed for consent, accountability, and transparency in AI tool usage within open source projects - Security: Enhanced vigilance required for AI-selected dependencies and AI-generated code review processes - Education: Comprehensive training programs needed for both AI tool users and community members affected by AI contributions - Funding: Increased resources needed for maintainer support and community health monitoring in AI-augmented environments ### 🧠 Learning Styles Assessment Reading level estimate: Grade 12-13 (technical presentation format) #### 👀 Visual learners - Present: Slides with frameworks and categorized information - Improve: Architecture diagrams showing AI agent interactions, flowcharts for decision processes #### 🔊 Aural learners - Present: Video presentation with verbal explanations and discussion - Improve: Podcast-style summary, guided discussion questions #### 🗣️ Verbal learners - Present: Clear definitions of AI alignment, analogies to existing OSPO work - Improve: Glossary of terms, story-driven examples of AI impacts #### 🧩 Kinesthetic learners - Present: Discussion format with interaction between presenter and audience - Improve: Hands-on workshops for implementing AI governance policies #### 🧠 Logical learners - Present: Structured framework covering use/release/contribute workflows - Improve: Decision trees for AI tool approval, metrics dashboards for tracking impact #### 🧑‍🤝‍🧑 Social learners - Present: Working group collaboration, community discussion format - Improve: Peer review processes for AI policies, collaborative policy development sessions #### 🧘‍♀️ Solitary learners - Present: Comprehensive slide deck for individual review - Improve: Self-assessment tools for AI readiness, individual reflection prompts #### 🌿 Naturalistic learners - Present: Discussion of community ecosystem health and sustainability - Improve: Analogies to natural ecosystem balance, lifecycle thinking about AI impacts ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Video format with spoken content - Improve: Detailed slide transcriptions, structured headings for navigation #### Keyboard navigation - --- ## Full Transcript [00:02] Okay, cool. And uh for our next agenda [00:06] item today, we have Emma Irwin, who will [00:08] be talking about AI and what OSPO should [00:11] be thinking about them. And I'll pass it [00:13] on to you, Emma. [00:15] >> Thank you, Natalyia. So, some [00:17] disclaimers about that. Um actually, [00:20] maybe I'll introduce myself first for [00:21] those that I haven't met before. I see [00:23] Remy, who I know it's always nice to see [00:25] you. Um who am I? So I've been working [00:28] in open source for 20 years. Um for for [00:32] 10 of those in the capacity of a [00:33] developer and for the last 10 or so um [00:36] sort of when you reach 20 you just keep [00:38] saying 20. Um as kind of a technical [00:42] program manager in the capacity of [00:43] helping organizations [00:45] be efficient and respectful and show up [00:49] in the open source ecosystem. So I [00:50] worked at Misilla for a number of years [00:52] on the open innovation team in what was [00:55] essentially an official OSPO at the time [00:57] just helping uh teams get good at open [01:00] source but also um innovating on a lot [01:03] of things including safety and digital [01:05] spaces and then uh for 5 years on [01:08] Microsoft's open source programs office. [01:10] So I was laid off there actually last [01:12] year um as many folks are. So, um I'm [01:17] not the disclaimer is more that I'm not [01:19] currently in an OSPO, but I've kind of [01:21] spent a lot of my time um kind of [01:24] figuring out and kind of in a way uh [01:27] fortunate to have a bit of time to just [01:29] start tinkering and and um I launched [01:32] the AI [01:34] um al AI and open-source alignment [01:39] working group in chaos and that came [01:41] from [01:43] uh so I watched a podcast or listened to [01:45] a podcast as many of us do and they were [01:47] talking about this topic of AI alignment [01:49] which is also known as AI safety. I [01:51] actually have a definition on one of [01:52] these slides but um and that was the [01:56] really compelling thing that came out of [01:57] that for me is that alignment is really [02:00] about ensuring that what AI systems do [02:02] is what humans want and need essentially [02:06] and that's also what when you hear open [02:08] or when you hear the word safety or AI [02:11] safety essentially that means the same [02:12] thing as well. Um yeah, does it do what [02:15] humans expect? Now there's like another [02:18] layer to that which is which humans [02:22] right and so um so a lot of what we do [02:25] in the AI alignment group um right now [02:28] we all well we just ran a s survey [02:30] actually maybe I'll pop it into the chat [02:31] a little bit later. I know that we're [02:33] closing it today or tomorrow, so I don't [02:35] know whether it'll work, but we're [02:37] trying to understand how AI is turning [02:39] up in people's uh lives and workflows [02:42] and businesses so that we can start to [02:44] form an opinion or maybe define a set of [02:47] principles [02:48] um that state these are the things that [02:52] that these are the principles of [02:53] alignment for open source and that that [02:56] represents like decades of labor, um the [02:59] community values as well as the business [03:02] value. So we're, you know, we want to [03:04] make sure that there's kind of this when [03:07] we ask the question of which humans, [03:09] there's always being this kind of [03:11] unofficial agreement between communities [03:14] and businesses that open source is both [03:16] of us, but there might be need to be a [03:18] bit more clarity around that. So that's [03:21] what and we did do a ton of sharing um [03:24] articles and we are building resources. [03:27] So if you join the AI working group just [03:29] to to view the resources that people are [03:31] always sharing then you might find it [03:33] worth it and then if you get around to [03:34] contributing or attending call that's [03:36] just a bonus but I like a lot of people [03:38] sharing things. So okay so that's my [03:40] preamble. Um I am working I am kind of [03:45] working for myself right now and I have [03:47] like some contracts but I'm not part of [03:49] any OS hospital right now. So, uh, there [03:51] might be a gap in what I'm thinking and [03:53] what hospitals are actually doing, but [03:55] maybe that's also superpower. I'm [03:57] embracing the I'm embracing the um the [04:00] wildness of of the world right now, the [04:03] good parts, I guess. So, I said some of [04:05] this already, and this is a lot of the [04:08] content in my site is generated from [04:10] blog posts I've written or uh other [04:12] things. So, um, I mentioned what [04:15] alignment is. Alignment is designing [04:17] systems so that their goals, behaviors [04:18] and uh decisions are consistent with [04:20] human values and intentions [04:23] in the context of chaos which is where [04:25] we are. Open source alignment means [04:27] consistent with what we've built known [04:28] value and expect um our metrics [04:31] specifically but also with the knowledge [04:32] that those metrics might need to change [04:34] to adapt to to that. So um and then I [04:38] also like to call out there's these two [04:39] contracts between communities in in open [04:43] source and businesses in open source [04:46] ospos like you're probably used to to [04:48] context switching. You know when you [04:50] talk to executives you talk about um you [04:52] know innovation and um reduced um like [04:57] capac like being able to to leverage the [04:59] community's um efforts and the hiring [05:02] pipeline. These are all things that [05:03] we've been saying to businesses and then [05:05] we when we talk to communities and folks [05:07] who you know are building those then we [05:09] talk about um you know making sure that [05:13] we're good citizens in the the community [05:15] and all these sorts of things. We've [05:16] always had those kind of two contracts [05:19] essentially. Um but my worry is a little [05:21] bit it's getting a little bit too [05:23] further apart the more we remove humans [05:25] from that. And so I won't go too much [05:27] into that but that's sometimes you'll [05:29] hear me refer to the two contexts or two [05:31] languages. Um and yeah I won't read the [05:34] rest of the slide but I think OSPO's [05:36] role if I was to be so bold as ensure [05:38] that alignment with humans remains a [05:39] priority as AI changes how open source [05:41] is built released and sustained. So, um, [05:45] if I was in an hospital right now, I [05:47] would be trying to I would be trying to [05:49] keep that, uh, central still because I [05:51] know that it's all about going fast [05:53] right now and making the money and [05:55] everything. And, you know, it's that's a [05:57] very easy thing to lose in this moment. [05:59] So, I also thought like I have these [06:00] three sections, use, release, and [06:03] contribute, which usually map to our OSA [06:05] priorities. And I have like just a [06:07] couple slides that synthesize some of [06:09] the things we've been talking about in [06:10] the AI alignment group. But then I [06:12] thought I'd pause in each of these [06:13] sections to just kind of hear from each [06:15] of you what you think the the problems [06:20] and work to be done etc is under each of [06:22] those if that sounds okay then I don't [06:23] have to just listen to myself talk. [06:27] So yeah context you know AI is now in [06:31] the middle of how we use open source. So [06:33] under the the using open source, here's [06:35] some just various data from some of the [06:37] sources that we've been sharing around [06:39] in our group which you might have also [06:41] seen out there. Um you know from a risk [06:44] perspective and what can go wrong you [06:46] know where we're where most OSOs have [06:49] tools for things like um typo squatting. [06:53] I don't know actually where the term [06:54] slop squatting comes but there's you [06:56] know just kind of a leveling up of how [06:59] uh there's both the hallucination of [07:00] names now coming into play and also you [07:03] know the oh yeah I'll share my sure I'll [07:06] share these after sure with a disclaimer [07:08] that you know um I put these together [07:11] pretty fast [07:14] uh it's really hard to understand [07:16] licensing in LLM so I actually spent a [07:20] bunch of time the last couple weeks even [07:21] with a full open source model. So that's [07:26] open model, open data, open everything. [07:28] And I found it almost impossible still [07:31] without like a lot of effort and maybe [07:32] some tokens to figure out what the [07:33] license is of data that something was [07:36] trained on. So this is really it's [07:39] really hard. Um there's um you know we [07:43] don't really know how some code is being [07:45] put together, what the context agents [07:47] are being given and so on. So I won't [07:49] read all of that. you know, there's 4% [07:51] of Claude commits are now a authored by [07:55] Claude alone. So, there's all this stuff [07:56] that's happening really, really fast. [07:58] And I and so I think that's really one [08:00] of the reasons I was glad to come today [08:02] is because I think we need to talk more [08:05] um [08:07] potential and this is where I'll like [08:10] ask you. I'm sorry there's so many words [08:11] on this, but I had to put it together [08:13] quickly. Uh potential open source [08:15] focuses OSO focuses. So I mean I I [08:18] personally always feel like an OSO's [08:20] role is very much about education of [08:23] engineers and um but maybe it's also [08:25] becoming education via two agents and AI [08:29] like this context setting for the AI [08:31] we're using I don't know but um so [08:34] there's a lot of different things that [08:36] like there's these three sections govern [08:37] dependency shape how agents operate and [08:40] protect the dependency ecosystem. [08:42] There's a lot of different work that we [08:44] could be doing. Um I'll just call out a [08:46] couple here. So um um [08:51] visible dependencies disappear. So [08:54] monitor the health of Yeah. So I'm I'll [08:56] just kind of over on this protect the [08:58] dependency ecosystem square like fourth [09:00] down there this monitor because this is [09:02] relevant to chaos monitor the health of [09:04] critical dependencies. I mean in the [09:07] past we have a chaos the bus factor [09:10] which became I think the contributors [09:11] factor and there's also response time [09:14] and all these sorts of things that we [09:15] have in the past thought about the [09:17] health of open source communities which [09:19] might themselves be changing so burnout [09:21] signals there we actually might see an [09:23] increase in merges we actually might see [09:26] more activity which makes it seem like [09:28] there's more contributors when it's [09:29] really AI and so I think there's some [09:32] adjustment that we need to do when we're [09:33] not just that we're monitoring the [09:35] health of our dependent ies, but looking [09:37] for s like trying to get through the [09:40] noise of the trying to find the human in [09:42] the middle and see like is are they [09:44] there? Are they are they actually [09:46] healthy or burnt out? So, um I think [09:48] there's definitely some innovation for [09:50] OSOs to do around monitoring the health [09:52] of their dependencies. [09:54] So, that's one uh I'll I'll choose one [09:56] from each square so that I don't uh so [09:59] going backwards [10:02] the shape how agents operate. So, you [10:05] know, this is a really interesting one [10:07] for me because I mean, I don't know how [10:09] much work you've all done building [10:11] agents. I've done like very little, but [10:13] I have done I have been working a lot [10:15] with coding like co-pilot and claude and [10:20] worked a little bit with agents and I've [10:22] I've realized that for example opening [10:25] is is in the training data. So if I ask [10:29] tell an agent to you know build an open [10:31] source solution it'll immediately go for [10:33] llama or like some other like well-known [10:36] models which actually are mixed license [10:38] and you know the training data is not [10:40] open and that kind of thing. So I think [10:42] that the context setting for the agents [10:43] and the AI we're working with coming [10:46] from hospitals will be really important [10:48] so the engineers for example are you [10:50] know have um that context provided to [10:54] them that they should be working with [10:55] initially to kind of cut through any [10:57] biases or hallucinations or or whatever [11:00] that might be happening with their [11:01] coding agents. Um [11:04] I also kn yeah I'll leave that for a [11:06] second. in governing governance. Um [11:11] there's a few things here. I mean these [11:14] are all exper I see these more as [11:15] experiments than like this is a thing [11:17] you should do. I think a OSPO should [11:19] experiment with context standards for [11:22] agents. They should experiment with [11:24] different um health models of [11:25] dependencies license compliance. Um [11:31] yeah I guess tracking tools published [11:33] and approved. Yeah. So there's there's a [11:35] few different governance things. I you [11:37] know I'm one of the things I'm doing [11:38] right now is working with the internet [11:40] society on upgrading their code of [11:42] conduct and we're adding an AI clause [11:44] around consent. So um that there's um [11:50] that anytime one of our maintainers [11:55] you know uses AI is with the consent of [11:57] the community anytime [11:59] uh we contribute upstream it's you know [12:01] there's a clarity and consent that that [12:03] project wants that. So I think that [12:04] that's also a governance thing. So those [12:07] are just some ideas. Um I have some more [12:10] questions we don't have answers to yet [12:11] and I think that's most of what we're [12:13] doing right now. So [12:15] the first one is actually based on a [12:17] project that I know is happening with [12:18] ecosystems uh right now which is should [12:20] agents manage abandoned dependencies. [12:23] Um, and you know, Andrew, I'm not sure [12:27] if you know them, is like incredibly [12:29] brilliant. And I've seen, you know, this [12:32] the idea that we could solve some of our [12:34] s sustainability and risk issues by, you [12:36] know, assigning angels to the job of [12:38] maintaining some abandoned or you [12:41] project. So, but is that something we [12:43] want to happen? What what does that look [12:45] like? Who decides? Are we working with [12:47] organizations? Is it just like if I was [12:49] at Microsoft, is this just Microsoft, [12:51] you know, with their agents in a version [12:53] or is this all like companies coming [12:56] together to do it? I I mean I don't [12:57] know. I think a lot of people are [12:59] talking about this kind of thing, but um [13:01] of course there's if we think about [13:02] alignment again, which I think is the [13:04] center of this is that alignment with [13:06] what communities, you know, who is the [13:09] community involved in that? And I think [13:11] there's just a lot of like removing [13:13] humans entirely is a risk. I can see [13:15] from a business perspective why that [13:17] would be desirable. [13:19] Who is responsible when agent selects a [13:21] dependency that turns out to be [13:22] malicious? I mean just generally the [13:24] ownership of AI everything is that the [13:26] engineer um like how does OSPO build [13:29] policy around that to protect them as [13:32] well because you know I'm I'm sure that [13:34] they're not you know no one wants to get [13:36] in trouble for something that they you [13:38] know do they blame AI and what's the [13:39] accountability there [13:42] um [13:45] yeah what should the if there's a [13:46] context library what should it contain I [13:48] mean there's actually like a ton of work [13:50] for OSOs now when I was laid off it felt [13:51] like ospos are being absorbed into [13:53] organizations because it was primarily [13:54] an engineering capacity but I feel like [13:57] there's just so much work cut out for us [13:58] at this point. Um at what point does AI [14:01] tooling require disclosure to your user? [14:04] This is again about consent. I think [14:06] consent is really central but um those [14:08] are some questions that I had. So maybe [14:10] I'll just pause there and you know just [14:14] among us how is AI turning up in your [14:17] like what questions do you are you [14:20] seeing or work that you're thinking [14:22] about in the use like taking a [14:24] dependency [14:30] >> well I can't I can't hear you Natalia [14:37] no I'm just thinking but [14:38] >> yeah I guess like one of the things I've [14:42] been thinking about um a lot of my work [14:45] um is centered around like repository [14:47] hygiene and repository standardization. [14:50] So [14:52] >> and then um [14:54] recently some of our repos uh shipped [14:57] like a agents.mmd to be able to [15:00] facilitate um work with um agents doing [15:05] development on their repo. So, I'm just [15:07] thinking of things like that, like how [15:09] can we set up the repo with enough [15:12] documentation [15:14] um [15:15] in that we can help guide agents to [15:18] doing development the way that aligns [15:22] with how we'd like them to do it. Um [15:25] yeah, so kind of thinking about like [15:26] setting the infrastructure and the [15:28] foundation [15:30] um to be able to support agents doing [15:32] work on repositories. Yeah. Yeah. [15:37] Interesting. [15:41] Anyone else thinking about anything [15:42] specific in use as it relates to AI? [15:46] I I don't know. But uh in terms of um so [15:49] what what we try to do in in in our case [15:52] is to kind of stay off all of the [15:55] security and and responsibility part and [15:58] just build recommendations and best [15:59] practices. So also approved licenses. [16:02] Yes. Okay. we have a kind of a list of [16:04] licenses that we recommend but then uh [16:08] for for everything else we just refer to [16:11] the to to security but yeah but but but [16:14] those are interesting questions that are [16:18] uh I think the the interesting one for [16:20] us is the around the dependencies and uh [16:25] because already just to understand what [16:27] your dependencies are it's a good [16:30] question and uh [16:34] Yeah. And I think there there's probably [16:37] somewhere in executive suites asking can [16:39] we just for for all these inactive can [16:42] we just fork them and maintain them or [16:44] rewrite them? I'm sure there's all kinds [16:46] of questions like that and you know the [16:48] hospital is going to be pulled into [16:49] those conversations I sus suspect. [16:53] >> Yeah. Because if you rewrite them then [16:55] probably someone else wants to rely on [16:57] them. I don't know. And then you start [16:59] to be provider of of the famous uh [17:04] um XKCD [17:06] right [17:07] >> leg of the whole thing but [17:14] >> okay I don't want to cut anyone off [17:15] though if anyone had any last things to [17:17] say about use but um [17:19] yeah no one's figured it out that's the [17:22] the one comforting thing in this moment [17:23] although it's also comforting when [17:25] things are figured out but it means [17:26] there's work to be done which is helpful [17:28] in this economy so release. So releasing [17:31] open source you know this is um [17:35] you know when I at Microsoft we the we [17:38] had policy and [17:42] sort of automation around releasing we [17:45] uh I mean I'm sure they still do but [17:48] like some of the things that we did [17:49] before AI was a factor was you know you [17:52] couldn't your repo couldn't turn to [17:54] public till you had sort of the business [17:56] sign off and the legal sign off and [17:58] those sorts of things. I think that kind [17:59] of process will still be important, but [18:01] there's there's kind of a lot more [18:03] that's coming into play right now. And [18:05] also the training of maintainers that [18:06] you just can't throw an open project [18:08] over the wall. That's probably more true [18:11] than ever. Um, but one of the [18:13] interesting things that I keep getting [18:15] stuck on too is like the open source AI [18:17] and I I feel like this is an area that's [18:20] because there's such a race right now, [18:22] there's there's just a lot of noise [18:23] around what that means. So for me just [18:26] making sure the engineers understand [18:28] what that means. So this OSI definition [18:30] which uh you know has the training data [18:33] can be doesn't have to be [18:37] I'm trying to avoid the word hidden but [18:39] you know um you don't have to open the [18:41] training data it can be described um [18:44] there's the meta the meta lama style [18:47] there's um vendor marketing there I mean [18:51] there's just a lot of things and when I [18:52] have gone into my my chat agent and said [18:55] you know what open- source AI [18:58] solutions solions do you have? I have to [18:59] really keep reprompting it and [19:02] reprompting it and reprompting it till I [19:03] get to something that I want. So I feel [19:06] like there's still a TBD around what [19:07] open source AI is right now like that we [19:10] actually understand what the license are [19:12] that we can reproduce the training that [19:14] there's community governance and no one [19:17] has achieved this and I don't even think [19:18] anyone's working on it to be honest but [19:21] um and maybe that this is me being like [19:24] an open- source kind of [19:27] I think they call us purists when [19:28] they're trying to insult but but I I I I [19:31] feel like this is work to be done as [19:33] well like you you know, um, [19:35] organizations are going to be have an [19:38] opportunity to lead how we talk about [19:39] open source AI. Oh my gosh, it's such a [19:42] bright screen. Potential open-source AIS [19:44] to do. So, um, [19:48] yeah, I just mentioned kind of the same [19:49] rules like of releasing, but maybe a [19:52] little harder to follow with AI writing [19:54] things. You know, I'm not sure how [19:56] hospitals are thinking about releasing [19:58] code that includes code that is written [20:00] by AI, which may or may not have the [20:03] right license. I know everyone's kind of [20:04] covering their eyes a bit right now. I [20:06] don't mean ospose but the world. Um [20:10] I mentioned the code of conduct needing [20:12] AI clauses in you know just from um if [20:16] if you're engineers releasing an open [20:18] source project I'd want to make sure [20:19] that my code of conduct has these type [20:22] of clauses and I'll give you an example [20:23] over at the internet society. I was in a [20:25] community call and there's like a ton of [20:26] AI notetakers suddenly just came into [20:29] play. and like wa it's like flies in the [20:31] room, right? And so, um, just making [20:33] sure there's the consent pieces and [20:35] interaction with your with your [20:36] community, making sure that engineers [20:38] understand how nervous essentially some [20:41] folks might be to these types of things. [20:43] And that can really help build the [20:44] reputation for your organization in [20:47] showing respect for individuals, the [20:49] humans that come in, the alignment [20:50] piece, [20:52] um, training, data, transparency. I [20:55] mean, OSO have always pushed [20:57] organizations to be more open. And I [20:58] think that includes things like data [21:00] transparency and getting just better at [21:02] at um helping people feel confident in [21:05] what they're using. Agents acting in [21:07] your name, self-explanatory. Um who's [21:10] accountable if you have agents merging [21:13] pull requests and you know if people [21:15] feel frustrated by that, who's [21:17] accountable for those frustrations? [21:20] Um [21:23] it's very easy. I I've had uh coding [21:25] agents want to put my secret into like a [21:28] file in that's not in git ignore and [21:31] there's all kinds of risk there. I'm [21:32] sure AI is getting really a lot better [21:34] at this but it doesn't mean engineering [21:36] engineers are prompting correctly. So [21:38] there's a lot of risk. I I heard you say [21:40] that you're not working your OSO isn't [21:43] working so much on security which kind [21:44] of makes sense that that's a a team that [21:46] works with OSO rather than OPO's job but [21:50] it depends on how big the organization [21:51] is I guess. [21:53] And then again, yeah, just community [21:55] plan, not just code. Uh, I think that [21:57] continues to be a risk for organizations [22:00] to make sure that their engineering [22:02] teams and I guess as AI becomes more [22:05] makes coding more accessible, there's a [22:07] whole new generation of people to teach [22:10] what it means to release an open source [22:12] project because it's not it's in a way [22:14] it's made releasing things more [22:16] accessible to another other generations [22:19] of people. So there's maybe even more [22:22] education to do there. Um, so yeah, some [22:25] questions and again I threw these [22:27] together so I hope they're relatively [22:29] good. Similar questions about agents [22:33] committing code on your behalf. Um, [22:36] yeah, your license, is it relevant when [22:38] you're [22:40] um, [22:42] if if it's just an agent and not a [22:44] maintainer, is is that okay for your [22:45] organization? Do you want to ask for a [22:48] specific commitment from your engineers [22:50] to turn up to to respond to any types of [22:53] questions? I mean, what does it mean to [22:55] interact with the activity when it when [22:58] you're lean starting to lean heavily and [23:00] your executives are leaning heavily on [23:02] you to do it? So, that that's like quite [23:05] challenging. And then should codes of [23:06] conduct. Again, I keep bringing this up. [23:08] I think because that's work that I'm [23:09] doing right now, but um I think that [23:12] that human piece is really really [23:14] important and probably the thing we're [23:15] most at risk in losing. And um in our AI [23:18] alignment call yesterday [23:20] um we have Coraline uh Mki from the [23:23] contributors covenant and ethical [23:25] licensing and she's really interested in [23:28] this you know not losing the people [23:30] part. Um and so that's why I emphasize [23:34] there as well. Anyways, I'll stop there. [23:35] So, releasing. So, your company putting [23:37] code out there. What questions do you [23:40] have or what are you running into? Any [23:42] inspiration and cool things that you [23:44] know about? [23:51] Well, I understand that for question [23:54] number two, [23:56] the so that the [23:59] scholars and in general practitioners [24:01] are converging towards um the fact that [24:05] uh that AI generated stuff is [24:08] non-copyrightable, right? [24:10] >> Mhm. [24:12] >> And [24:15] so it's kind of Yeah. And so I don't [24:16] know. Uh [24:19] and [24:21] then the license. Uh [24:28] yeah, there was an article someone had [24:30] posted in our AI working group about [24:33] licenses. Let's see if I can quickly [24:35] find it. [24:38] Um fair. Yeah, it was called fair. I [24:40] don't know anything about fair source or [24:42] like how it's being described, but this [24:45] was something that came into our chat [24:47] the other day. Maybe it's helpful. [24:52] And and um [24:55] >> something that I haven't really called [24:56] out here is the cognitive load. Um [25:00] there's an article on co Let me just [25:02] find that too. Hold on a second. Um [25:10] I'll try and find it when I'm you're not [25:12] having to watch me scroll through it. [25:13] But [25:14] >> yeah, [25:15] >> there's already the the discovery of [25:17] cognitive overload for open source [25:19] maintainers and there's the pressure [25:22] from executives to, you know, go faster, [25:24] use AI, these types of things. What is [25:28] the OSO's role in [25:31] setting helping engineers set boundaries [25:33] for their cognitive load? I I feel like [25:35] that's a thing as well in the future. Um [25:38] yeah of copyright is yeah I saw that [25:43] copy left is dead. [25:46] Yeah, one thing that um our ospo has [25:49] been thinking about, I had a [25:51] conversation with Remy [25:54] about this um for assisting maintainers [25:57] with I guess like the burden of having [26:00] to go through so many PRs now that um [26:04] lots of people are using AI to make [26:07] contributions to repositories. Oh, the [26:09] formatting of this is not the greatest, [26:12] but we were thinking of kind of setting [26:15] up uh contributing guidelines that um [26:18] guide the way that repositories accept [26:21] PRs. So for instance, we don't accept [26:24] PRs that don't directly address an [26:27] existing issue, [26:29] >> touch more than a certain amount of [26:31] files, contain more than a certain [26:33] amount of commits, contain uh change [26:35] core system files, are untested, don't [26:38] pass a test suite. And um these are [26:41] guidelines that repositories can set [26:43] themselves. And they can also just [26:45] include that oh we are just not [26:47] accepting [26:48] um repositories that are entirely AI [26:51] generated. But just having these like [26:53] rules and restricting the [26:57] um restricting um or adding certain [27:00] limitations to how PRs are um structured [27:04] to ensure certain things. So for [27:08] instance, if we limit the amount of like [27:10] files and commits, it would reduce the [27:13] amount of review time needed from a [27:15] maintainer or from a human and kind of [27:17] reduces like the radius, right? Um [27:21] testing suite just ensures there's no [27:22] breaking changes. Um ensuring that PRs [27:26] address existing issues means that it [27:30] reduces the risk of I think the word was [27:33] roadmap drift because it ensures that [27:36] contributions made are align with like [27:39] um the road map and the priorities of [27:42] the project. So just having these [27:43] certain guidelines can help the way that [27:46] we review PRs and accept issues as well. [27:50] But that's one thing our OPA has been [27:52] thinking about. Um we haven't put this [27:54] into practice quite [27:56] >> quite yet and um haven't or are unable [28:00] to talk about like actual tangible [28:01] results but this is one of the ways we [28:04] were thinking of guiding um AI [28:06] development in repositories. Yeah. And [28:08] and did you actually do it like [28:10] proactively before like the problem [28:13] arised? And uh [28:14] >> you can totally do that too. Yeah. [28:17] >> Okay. [28:19] I love that roadmap drift. Is that what [28:21] you said? [28:22] >> Yes, that was the I think. [28:23] >> Yeah. Yeah. It's like and and in [28:27] technology we know that neurodeiverse [28:29] people are over represented. So actually [28:30] a lot of us are actually quite good at [28:32] jumping between a lot of things. Um but [28:34] you know that can actually be doubly bad [28:36] as we're sort of like following the [28:38] interest of um what people are [28:40] contributing. I love the the idea of [28:42] staying on track there. [28:43] >> So just to go to contributing here [28:45] because it's sort of our our um the uh [28:49] runway home here. This is just what we [28:52] were just saying, right? Like there's [28:54] just a lot of the maintainer crisis is [28:56] getting worse. Um, [28:59] and I actually have like a little graph [29:01] here to show how folks are we have a a [29:04] read me in the AI alignment working [29:06] group repo which I'll try and share um [29:07] after this where we're we're trying to [29:09] track LLM [29:11] um projects but um you know it it's a [29:15] whole layer on top of the already [29:17] critical open- source sustainability [29:21] issue and um [29:25] so yeah it's causing obvious obviously a [29:28] lot more problems and there's lots of [29:29] way that that organizations are trying [29:31] yeah that is an yeah ours is um ours is [29:35] actually listed on that list so that [29:38] would be so I think there's this like [29:41] how open source projects are responding [29:43] but also recognizing the sustainability [29:45] risk something I did when I was laid off [29:47] is like worked more in the [29:49] sustainability space and I just I feel [29:51] like [29:52] I mean and I know from talking to [29:54] Android ecosystems that there's fewer [29:57] work. [30:06] >> I think I lost the audio. [30:10] >> Yes, I don't think we can hear you, [30:11] Emma. [30:16] Let me um [30:18] >> Oh, now we do. [30:19] >> Okay. Yeah, this is the problem. Um I [30:21] apologize. [30:23] I think it might have something to do [30:25] with Firefox because as soon as I open [30:26] Firefox this started happening. Um I was [30:30] just what I was trying to say was um [30:35] working in the open source [30:36] sustainability. [30:38] So this is becoming more prevalent. The [30:40] exhaustion is becoming more prevalent. [30:43] There's fewer and fewer [30:46] companies contributing to open source [30:48] financially and that's backed up by some [30:50] of the data and ecosystems. they can [30:52] track the GitHub sponsors number [30:55] uh like whether a project is sponsored [30:56] or not. And I don't think Andrew's [30:58] shared those numbers publicly yet, but [31:00] they're, you know, available in their [31:01] data sets that it's decreased. It's just [31:03] going down. There's fewer maintainers [31:05] being sponsored, at least through GitHub [31:07] sponsors, and I think actually um [31:11] Open Collective is also reporting less [31:14] less less funding. um the rallying cry [31:17] to do more about this, the rallying cry [31:19] to do more about to do more funding is [31:21] louder than ever, but the actual actions [31:23] do not seem to be translating out of [31:26] that. And so I would suggest that it's [31:28] also an hospital's responsibility to [31:31] help their executives see that as a as a [31:36] opportunity to show up in the ecosystem [31:38] in some way. Um Foss Fund was a program [31:41] I ran at Microsoft and there's still I [31:42] think Foss funds being run but I think [31:44] there's a 2.0 0 of that where we you [31:47] know you apply metrics for chaos you [31:49] look for maintainers who are struggling [31:50] under this kind of thing help them set [31:52] up governance not just give money but [31:54] give time and help them set up I think [31:57] that there's an opportunity there [31:59] anyways um off my platform there um [32:02] potential to-dos yeah upstream [32:05] contribution policy protect your [32:06] dependencies attribution [32:10] uh I think is something emerging out of [32:13] um code uh AI I type of um efforts. Uh [32:18] it was interesting. Coralina who's in [32:20] our group used to work at GitHub like [32:23] many many years ago and created an [32:24] action that allows you to ban [32:28] user or to ban contributions from users [32:30] that you have banned from the [32:31] organization. And so people are banning [32:33] Claude [32:35] and they're b so she's like this tool I [32:37] made 10 years ago is suddenly like new [32:39] again. So um [32:43] so there's all types of things that [32:44] people are thinking about um the [32:46] authorship I guess of participating in [32:48] ecosystem norms is it's a massive [32:50] opportunity honestly to show to help [32:53] companies show up right now uh for [32:55] communities and for people for [32:57] maintainers for um you know for chaos to [33:00] to build new metrics that kind of build [33:03] on the work that we've done. It's not [33:04] like a lot of the metrics have gone [33:06] away, but there's a new layer to them. [33:07] And I think that, you know, helping [33:09] people evaluate their ecosystems better. [33:12] Um, these are just some more questions. [33:15] They're they're kind of similar. [33:29] >> I think we lost my gain. [33:32] >> Yes. in. [33:34] >> Yeah, sorry. I don't know what's going [33:36] on. I'm almost done here. So, hopefully [33:38] I'll just I'll just have to put up with [33:40] it. There's there's something I need to [33:42] figure out that's going on with my [33:43] system. A reboot helped, but then when I [33:46] opened Firefox, it started again. So, I [33:47] think that might be my culprit, which is [33:49] weird for internet, but [33:52] and then um if you're if there was a [33:56] action framework, I was just playing [33:57] with this idea. Can you hear me now? [34:00] Yeah. is I mean I think it's always OSPO [34:03] for me is education, teaching, [34:06] capacity building, capability, [34:08] competence building and just especially [34:10] I think as you're as more non-engineers [34:13] start to to work with open [34:16] um things it's becoming the hospital's [34:18] role is actually becoming more important [34:22] or I think they're sometimes called OS [34:23] AI project offices or whatever the [34:27] so these are some more framework work [34:30] ideas. I won't read them all because [34:31] we're almost out of time and my [34:32] connection is bad. [34:35] Um, I also have this slide which I'll [34:37] share my slides later in case they're [34:39] helpful. Um, [34:41] I have a bunch of the sources down below [34:43] too. So, um, [34:48] so yeah, these are just some of the [34:49] things that we've been talking about in [34:50] their AI alignment group too. So, [34:51] they're not just all Emma. Um, I guess [34:54] the last thing I would say is there are [34:57] experiments. Yeah, I mentioned this [34:59] already. I mentioned this already around [35:02] agents that maintain open source. [35:04] What does it mean for businesses and [35:06] business partnerships? I think that [35:07] partnership piece for OSOS is important [35:09] to figure out how you're going to work [35:10] with other companies on this. And yeah, [35:13] so what you should do for sure is join [35:15] our AI alignment working group chat like [35:18] just come and hang out with us. You [35:20] don't have to join calls or anything, [35:21] but there's lots of great links being [35:22] shared. We'd love you to share what [35:24] you're learning. I think it's all about [35:26] learning right now and all the rest of [35:29] the things on this slide. And that's [35:31] that's what I brought today. Um I'm sure [35:33] there's a lot more things we could say, [35:34] but that's that's kind of uh a few [35:36] things that maybe are helpful and I'd [35:38] love to hear continue the conversation. [35:46] Okay, awesome. Thank you so much, Emma, [35:48] for giving your presentation on AI and [35:51] OSOS. Um there's been a lot to think [35:54] about in this space. So it's really nice [35:56] to um see everything come together and [35:59] just have everything at least a lot more [36:01] organized in my head. Um yeah. So you I [36:07] saw that the uh AI alignment working [36:10] group has a survey out. Um, I feel like [36:13] some attendees in this uh call can [36:16] totally fill out the survey [36:21] if you can drop it in the chat. I think [36:23] I h I have in the agenda too actually. [36:25] Let me bring it up here. [36:33] Okay, so here's the survey by the [36:36] working group. Feel free to fill it out. [36:39] Um, let's see. And we have two minutes [36:41] left in our meeting. Any other final [36:43] thoughts that anyone would like to [36:45] share? [36:52] Okay, awesome. Um, okay. And in terms of [36:58] the future, next week we have a [37:01] presentation [37:03] by uh Travis Hathway who will be talking [37:05] about how he uses chaos at Anaconda. Um [37:08] so feel free to join us for that two [37:11] weeks from now and we will have some [37:14] other exciting presentations coming up [37:16] in the future. Yes. Otherwise I think [37:19] that's all for today. So I hope that [37:21] everyone has a great week. Thank you [37:24] Emma again. Thank you everyone for [37:26] >> for hosting. [37:28] >> Thank you. [37:31] >> Thank you Natalia. [37:32] >> Yes. Thanks Alan. [37:35] Bye [37:36] >> bye.  📄 **Full text:** ![[Full Document Contents/Articles/CHAOSS OSPO WG, March 19 2026.md|CHAOSS OSPO WG, March 19 2026 by YouTube]] ## Highlights - value. So we're, you know, we want to make sure that there's kind of this when we ask the question of which humans, there's always being this kind of unofficial agreement between communities and businesses that open source is both of us, but there might be need to be a bit more clarity around that. So that's what and we did do a ton of sharing um articles and we are building resources. So if you join the AI working group just to to view the resources that people are ([View Highlight](https://read.readwise.io/read/01kpb5be67bgvvqs7hsyn68jaa)) - Tags: 👻 ai highlighted - of the reasons I was glad to come today is because I think we need to talk more um potential and this is where I'll like ask you. I'm sorry there's so many words on this, but I had to put it together quickly. Uh potential open source focuses OSO focuses. So I mean I I personally always feel like an OSO's role is very much about education of engineers and um but maybe it's also becoming education via two agents and AI like this context setting for the AI ([View Highlight](https://read.readwise.io/read/01kpb5be7adh87hra7ema1he0f)) - Tags: 👻 ai highlighted - um to be able to support agents doing work on repositories. Yeah. Yeah. Interesting. Anyone else thinking about anything specific in use as it relates to AI? I I don't know. But uh in terms of um so what what we try to do in in in our case is to kind of stay off all of the security and and responsibility part and just build recommendations and best practices. So also approved licenses. ([View Highlight](https://read.readwise.io/read/01kpb5be95m496j6ayqxw5dxf3)) - Tags: 👻 ai highlighted - there's there's kind of a lot more that's coming into play right now. And also the training of maintainers that you just can't throw an open project over the wall. That's probably more true than ever. Um, but one of the interesting things that I keep getting stuck on too is like the open source AI and I I feel like this is an area that's because there's such a race right now, there's there's just a lot of noise around what that means. So for me just making sure the engineers understand what that means. So this OSI definition ([View Highlight](https://read.readwise.io/read/01kpb5becpezfj9csx4m0szpkm)) - Tags: 👻 ai highlighted - there's an article on co Let me just find that too. Hold on a second. Um I'll try and find it when I'm you're not having to watch me scroll through it. But >> yeah, >> there's already the the discovery of cognitive overload for open source maintainers and there's the pressure from executives to, you know, go faster, use AI, these types of things. What is the OSO's role in ([View Highlight](https://read.readwise.io/read/01kpb5behpkqfw5p888fabmw5n)) - Tags: 👻 ai highlighted - agents that maintain open source. What does it mean for businesses and business partnerships? I think that partnership piece for OSOS is important to figure out how you're going to work with other companies on this. And yeah, so what you should do for sure is join our AI alignment working group chat like just come and hang out with us. You don't have to join calls or anything, but there's lots of great links being shared. We'd love you to share what you're learning. I think it's all about learning right now and all the rest of the things on this slide. And that's ([View Highlight](https://read.readwise.io/read/01kpb5betekb75fj9x0ev0a6rj)) - Tags: 👻 ai highlighted