Singapore's Minister for Foreign Affairs, Dr Vivian Balakrishnan, gave an unusual keynote at AI Engineer Singapore on 16 May 2026. In front of 2,000-plus builders, he walked through the personal AI setup he runs entirely on a Raspberry Pi: Ollama to serve the models, whisper.cpp to transcribe his voice, NanoClaw as the agent layer, and Obsidian holding his notes underneath. No grand policy framing, no slides full of national statistics. The working stack on stage was the argument.
The point he was making: sovereignty without protectionism. If serious AI workflows run on cheap hardware with open-weight models — models anyone can download and inspect — then the local AI agenda is about owning the capability. Individuals, agencies and enterprises should be able to inspect, copy and modify their stack, not be told which foreign models they may use. Singapore's Ministry of Foreign Affairs published the full transcript.
The timing sharpened the message. On the same conference day, OpenAI and the Singapore government signed a S$300 million agreement at ATxSummit for OpenAI's first Applied AI Lab outside the United States, and Minister for Digital Development Josephine Teo's separate AI Engineer talk pushed in the same direction. A cabinet endorsement of local, open-source AI on the same day as a flagship deal with a US frontier lab is not a contradiction. It is the dual track Singapore has been signalling through IMDA and the National AI Strategy 2.0: anchor the frontier partnerships, and keep the option of fully self-hosted stacks for sensitive workloads.
Related resources
Go deeper on this topic
Vendor directories
Ready to move
Reader notes
Questions, corrections, and field notes
Curated notes from verified readers. Submissions are reviewed before publication.
Loading reader notes...

