SINGAPORE May 17, 2026 – In a striking demonstration of hands-on tech adoption, Singapore’s Minister for Foreign Affairs Dr. Vivian Balakrishnan has developed and deployed his own AI agent – a personalized “second brain” to support his demanding diplomatic role. The system, which he detailed in a recent speech and public technical write-up, runs on modest hardware and leverages open-source tools.
Speaking at the AI Engineer Singapore conference on May 16, 2026, Dr. Balakrishnan explained how he assembled NanoClaw, a self-hosted AI assistant built around Anthropic’s Claude model. The setup integrates with WhatsApp for messaging, processes voice notes via local transcription, handles images, and maintains a compounding knowledge graph drawn from his speeches, briefings, articles, and interactions.
“This is not rocket science,” he noted casually in a widely shared video clip from the event. The core hardware is a Raspberry Pi (with just 8GB RAM), running in Docker containers for isolation and security. It uses tools like Ollama for local embeddings, whisper.cpp for on-device voice processing, and concepts inspired by Andrej Karpathy’s LLM Wiki for synthesizing facts into browsable knowledge in Obsidian.
Key Features of the System
- Persistent Memory: Unlike stateless chatbots, it builds a graph database of facts, entities, and relationships that grows over time, enabling semantic recall during conversations.
- Workflow Integration: Communicates via WhatsApp, drafts speeches, prepares briefings, condenses information, and supports daily tasks. It even pulls from Gmail and can handle multimodal inputs.
- Security Focus: Sensitive data stays local; API keys are proxied, and per-group isolation prevents cross-contamination. Dr. Balakrishnan emphasized transparency, noting that even if compromised, it mostly holds publicly curated material.
- Accessibility: Built by “assembling tools” rather than heavy coding, highlighting how barriers to advanced AI have collapsed.
In his speech, Dr. Balakrishnan stressed a broader philosophy: leaders and professionals must experiment directly with technology rather than relying on briefings. “You cannot govern a technology you have only been briefed on,” he said. He advocated for decentralized, individual-level AI adoption to re-engineer workflows across sectors – from diplomacy to everyday professions – aligning with Singapore’s Smart Nation goals of broad deployment over frontier model development.
The minister shared the full architecture in a GitHub Gist, encouraging replication and further innovation. Reactions on social media have been enthusiastic, with many praising Singapore’s practical, ground-up approach to AI and contrasting it with more passive strategies elsewhere.
Dr. Balakrishnan, a former eye surgeon turned politician, continues to tinker with the system, which he says has become indispensable – to the point he runs parallel versions during updates and “has not dared to switch it off.”
This initiative underscores Singapore’s push to empower citizens and officials alike with AI, positioning the city-state as a leader not just in policy but in personal technological fluency. As Dr. Balakrishnan put it, the real value emerges “workflow by workflow, sector by sector, and at the individual level.”
About The Author
