OpenAI Joins the Linux Foundation's New Agentic AI Foundation (nerds.xyz) 18
OpenAI, alongside Anthropic and Block, have launched the Agentic AI Foundation under the Linux Foundation, describing it as a neutral home for standards as agentic systems move into real production. It may sound well-meaning, but Slashdot reader and NERDS.xyz founder BrianFagioli isn't buying the narrative. In a report for NERDS.xyz, Fagioli writes: Instead of opening models, training data, or anything that would meaningfully shift power toward the community, the companies involved are donating lightweight artifacts like AGENTS.md, MCP, and goose. They're useful, but they're also the safest, least threatening pieces of their ecosystem to "open." From where I sit, it looks like a strategic attempt to lock in influence over emerging standards before truly open projects get a chance to define the space. I see the entire move as smoke and mirrors.
With regulators paying closer attention and developer trust slipping, creating a Linux Foundation directed fund gives these companies convenient cover to say they're being transparent and collaborative. But nothing about this structure forces them to share anything substantial, and nothing about it changes the closed nature of their core technology. To me, it looks like Big Tech trying to set the rules of the game early, using the language of openness without actually embracing it. Slashdot readers have seen this pattern before, and this one feels no different.
With regulators paying closer attention and developer trust slipping, creating a Linux Foundation directed fund gives these companies convenient cover to say they're being transparent and collaborative. But nothing about this structure forces them to share anything substantial, and nothing about it changes the closed nature of their core technology. To me, it looks like Big Tech trying to set the rules of the game early, using the language of openness without actually embracing it. Slashdot readers have seen this pattern before, and this one feels no different.
Great (Score:2, Insightful)
Another puppet.
WTF (Score:5, Insightful)
The Linux Foundation has what now?
I know it's a corporate front group but this is a whole new level of fucked up shit. Fuck the Linux Foundation.
Re:WTF (Score:5, Informative)
The Linux Foundation has many large corporate members, including ones you wouldn't consider to be very open-sourcey.
https://www.linuxfoundation.or... [linuxfoundation.org]
Re:WTF (Score:4)
Re: (Score:1)
It's not like AI is going away. And Linux users will want to use AI. And in the best case it is open AI in an open AI ecosystem, so it is the best interest of the Linux foundation to organize a base others can build on to include the AI for users.
The WTF is not the Linux foundation. The WTF is what OpenAI will do in the gremium. Their interest is, other than the name suggests, closed AI.
Re: (Score:2)
A real open project supplies the preferred components for completely rebuilding the software. That means, you should get from such a project all the source documents (not preparsed token datasets) for training and all the training scripts they use themselves and all the specs. If you do
Re: (Score:2)
There are a few that provide you everything (look for example into Olmo), but you're no thinking it through when you want to build it from source like you do maybe with your browser. Building your own browser duplicates work (in particular if you use the same setup as the official builds), but it is negligible at modern computers. I were there, when compiling Mozilla took more than a day, people were very glad they were provided with ready-made binaries.
Now we're talking about AI models. The large companies
Re: (Score:2)
However, open source is not about giving out a model for cheap/free to whoever asks. It is about giving away the foundations that allow complete duplication, so that other members of humanity, smarter or more informed, can contribute and/or branch away from the work.
The cost of training is irrelevant. It merely reflects the
Re: (Score:2)
I think Olmo had clear dataset downloads. Need to check later.
I appreciate your response and think you've got a few good points, but I still don't see the training data to be the important one. Let's be honest, 80% of data is crap anyway. I think the commercial curated datasets have a higher standard, but if you take some crawl-dataset you first need good quality filters and then you notice that its outdated and you somehow should get at least a bit of recent data so people asking about the current presiden
Re: (Score:2)
No, absolutely fucking not.
AI may or may not go away, but it's at the "Everything will be a touchscreen!" stage of hype. Just as Windows 8 was a bad idea, so is trying to graft an "Agentic OS" onto a desktop OS. It's fucking stupid.
And AI isn't going to become open until the processing and storage requirements are within normal parameters, which is at least 10-20 years off, possibly longer given Moore's law is what it is. Some people are running it locally, but they're not getting responses that are quick,
Re: (Score:2)
Bad news, but everything *has* a touchscreen. The majority of daily used end-user devices are mobile phones. If people would have rejected adding touchscreen support (or vendors and Google wouldn't be allow to implement it), these phones would probably be Windows phones by now.
AI is already within "normal parameters", you can start with good networks at GPUs with 8 GB of RAM. If you have enough time you can use your raspberry pi. The Qwen3 4B model that fits the pi or smartphone is better (due to size at in
OpenAI isn't open? (Score:2)
Crap source (Score:2, Troll)
Some of their points are valid, but their writing is so toxic that they make a poor source of information.
Such glorious infrastructure! (Score:3)
Given that the whole thing is just an exercise in getting away with bots being more or less as OK-ish with poorly structured inputs as they are with anything else it's not like it would be a better 'standard' if there were a thicket of XML schema involved; but saying:
We’re committed to helping maintain and evolve this as an open format that benefits the entire developer community, regardless of which coding agent you use."
About a 'standard' which is 'put some kind of markdown, y'know, stuff you'd tell someone about your project in a text file called Agents.md' is a little grandiose.
Re: (Score:2)
It is rather more readable, because LLM like well-structured human-readable text more than syntatic dense Makefiles. It also has not that much to do with robots.txt which is a robot exclusion protocol.
Re: (Score:3)
Better an underengineered AGENTS.md than an overengineered MCP standard. If you compare MCP to UTCP, you'll see that the only point to MCP is to better integrate it with commercial pipelines. UTCP just defines a structured format to tell the LLM about the tools and then relies on the tool calls using standard protocols like HTTP requests. MCP requires you to implement a proxy executing the requests even when your inference engine already could execute the required tools without MCP.
The Agentic AI Foundation belongs (Score:2)
where the sun don't shine.
The Linux Foundation has always been kind of useless, but they're really outdoing themselves this time.