Written by Bernardino Realino Aryo Lintang.
Everyone Is Building Something: What AI Engineer Singapore Taught Me
6 min read · Part of Sponsoring students at AIE became a test of Singapore's AI ambition
Written by Bernardino Realino Aryo Lintang.
6 min read · Part of Sponsoring students at AIE became a test of Singapore's AI ambition
I almost didn't go. The ticket price was steep for a student covering household groceries on a tight budget. I spent days weighing it, thinking, " Is this worth it, or am I just chasing hype?” In the end, I applied for a scholarship slot, got in, and walked into Capitol Kempinski on 15 May, not entirely sure what to expect.
By the end of day one, I had my answer. This was not a conference about AI. It was a conference about builders, and the line between who gets to build and who doesn't has been thinner.
The room was full of people shipping things. What struck me first wasn't any particular talk. It was the energy. Every conversation I walked into, be it at the registration queue, between sessions, or during lunch, people were mid-sentence about something they were actively building. Not planning. Not researching. Building.
There's a version of the AI hype cycle that is mostly observers watching other observers. AIE was the opposite. The 1,000+ people in that room were engineers, founders, designers, and operators who had shipped something, broken something, and were figuring out what to build next. That energy is contagious in a way that no YouTube video or podcast can replicate. You leave sharper just from being around it.
One of the moments that will stay with me happened almost accidentally. My friends and I were introduced to Geoffrey Huntley, the open-source researcher and developer behind Ralph Loop, a project exploring what it means to build your own coding agent.
We introduced ourselves as students, half-expecting the usual polite encouragement. Instead, he looked at us and said something along the lines of: " Build your own coding agent. You can. Anyone can.”
He was direct about it: tools like Cursor, at their core, automate the loop of reading files, listing directories, editing code, and running commands in a terminal. That's the skeleton. What separates the tools people actually use isn't some proprietary magic, but it's hunger, iteration, and how the story gets told. The moat is marketing and momentum, not the underlying architecture.
That hit differently than I expected. I've been building AI tools for months, always with a sense that the serious infrastructure was being built somewhere else, by people more qualified. Geoffrey's point, delivered without fanfare, was that the gap between a student hacking on a weekend and a funded startup is smaller than the ecosystem would have you believe. Whether you close that gap comes down to whether you're hungry enough to keep building when it isn't working. I left that conversation with a project idea and a recalibrated sense of what's possible.
A Minister, a Raspberry Pi, and a Personal AI System
The moment that genuinely stunned me came from a different direction entirely.
Dr Vivian Balakrishnan, Singapore's Minister for Foreign Affairs and a former surgeon, shared how he built his own personal AI system, NanoClaw, running on a Raspberry Pi 5. Not commissioned to a vendor. Not deployed by a government agency and assembled by him.
The system is built around the cognitive demands of his role: as a foreign minister, he needs to hold in his head the histories, conflicts, cultural contexts, and interpersonal dynamics of dozens of countries simultaneously. Parliamentary questions need to be prepared. Briefings need to be synthesised. The cognitive load is enormous.
His solution was to build something. NanoClaw uses Baileys to connect to WhatsApp, Whisper for speech-to-text, and Obsidian as the interface layer. It has a memory system and an analysis system. The key phrase he used was one I keep coming back to: assembled, not written.
He wasn't talking about writing code from scratch. He was talking about the mindset shift that AI enables, the ability to compose, orchestrate, and connect tools into something that fits your specific life and workflow. He learnt by doing, iterating until it worked for him.
Prime Minister Lawrence Wong has reportedly vibecoded four applications for his personal use. This is not incidental. When the people leading the country are themselves building with AI, it sends a signal to everyone watching: this is not a spectator sport.
What moved me most about Dr Balakrishnan's story isn't the technology. It's the posture. Here is someone with decades of expertise and every resource available, choosing to understand the tools himself rather than delegate that understanding away. That's the model I want to carry forward.
What Singapore's Ecosystem Actually Needs
Watching all of this, I kept thinking about the students back at NUS who couldn't attend; not because they weren't interested, but because the price point and the timing made it inaccessible.
Here's what I think Singapore's ecosystem needs if it wants to stay at the frontier: AI literacy needs to move into classrooms before students have to seek it out themselves.
The pace of change makes this harder than it sounds. When I started my data science degree, the conversation was about LLMs and prompt engineering. Now it's multi-agent orchestration, MCP protocols, evaluation pipelines, and context engineering. The curriculum that was cutting-edge eighteen months ago is already catching up to where practitioners have moved on from.
Formal education moves slowly. The builders at AIE move fast. The gap between those two speeds is where Singaporean students get left behind, not because they lack talent, but because the structures around them haven't caught up.
The answer isn't just more AI modules in universities. It's building a culture where tinkering is expected, where shipping small things is celebrated, where the standard isn't "did you study AI" but "what have you built with it." Dr Balakrishnan on a Raspberry Pi is the more compelling curriculum than any lecture slide.
What I'm Bringing Back
I came to AIE with a list of questions about production systems, evals, and multi-agent architecture. I left with those questions partially answered and a longer list of new ones — which is exactly how learning is supposed to feel.
But beyond the technical takeaways, I'm carrying something harder to quantify. The conference reminded me that the most important difference between the people building the future and the people watching it isn't credentials, funding, or even raw skill. It's whether you're willing to sit down, assemble something from what's available, and keep going when it breaks.
Anyone can build. The question is whether you're hungry enough.