During the Open Source Summit Europe keynote, Jim Zemlin, executive director of the Linux Foundation, revisited the LLM landscape with an emphasis on the effect it has on the industry, the security, quantity and quality of the developed code. He emphasised that DeepSeek’s revolution was validated by the gpt-oss release, OpenAI’s first open-source model.
He began by reemphasising the significant societal and economic value that open-source has on the global community. Referring to Frank Nagel’s Harvard research paper, he underlines the 9 trillion dollars value that companies around the world “take for granted” when using open source. Zemlin also mentioned that Nagel joined the Linux Foundation as the advising chief economist, with the intention of continuing his research and expanding it to emerging fields like artificial intelligence and developing frameworks to support growth in the commercial OSS startup ecosystem.
Further, he highlighted the crossroads moment for open-source and AI with the release of DeepSeek’s open-source, open-weight model, demonstrating that such models can be “equally performant”. Zemnal attributes DeepSeek’s success to China’s decades-long bet on open-source, inspired by the US’s similar journey. This adoption is present in different parts of the Chinese economy, not only in the private sector but also in the government.
Liang Wenfeng: In the face of disruptive technology, the moat formed by closed-source systems is short-lived.
He stated that this direction is validated by OpenAI’s decision to release the first open-source model, joining a multitude of open-source models available on the market, which have already caught up in terms of performance with proprietary models.
Regardless of their evolution, frontier models are so “…2024, while in 2025 we are shifting towards agentic AI”. He believes that the true potential of AI will be discovered in this “agentic era”, as we will have tools that can perform sequential tasks for us—standards like MCP and A2A extending the frontier models’ abilities with the capabilities of accessing resources more efficiently.
Next he questioned the audience and provided several opinions from the media space related to the redundancy of human coders near future, continuing by sharing the results of different studies on the use of generative AI tools which concluded that coders don’t really trust the output of those tools, or even if they increase the individual’s productivity it affects the general productivity of the team in a bigger proportion.
He concluded his presentation by arguing that the open-source community should focus their efforts on embracing the AI as a potential productivity accelerator, but keeping in mind the need for maintainer oversight, security scanning, and transparency for AI-generated code. I this way, Zemlin thinks, the developers will not be replaced by AI coders but by “hyper-motivated”, “hyper-productive” human coders that use the best tools for the job.
