Singapore Field Notes: Curated Growth and the Next Wave of AI Diffusion
Kishore Mahbubani’s “Asian Century”, and on-the-ground talks with Tencent, ByteDance, Google, Meta and startups
Reflections on Singapore Trip
Hello, I’m back in Hong Kong now. Still processing what was a really wonderful trip to Singapore. It was my first time back since I left six years ago. In that time, a few major shifts have reshaped APAC in ways that all seem to flow through Singapore: the talent reshuffle from Beijing and Hong Kong, the wave of capital, and its growing role as a tech and financial hub in Asia. A lot of my friends have moved there from Beijing, Hong Kong, and even New York.
Walking through Duxton Hill, Orchard Road, Tanjong Pagar, and Dempsey Hill was a real walk down memory lane moment. The same mix is still there: colonial buildings, old military barracks turned into restaurants, highrises but covered in meticulously curated greenery, lanes of shophouses home to fortune tellers, coffee bars, and electronics shops. But what struck me most was how much the energy has changed since I last lived there.
When I lived in Singapore, the expat community still felt like it was dominated by Aussies due to the proximity and Brits due to the colonial history that worked for various FMCGs or pharma companies. But since then, Brexit has driven even some of the most successful British companies, such as Dyson, to relocate to the tropical city, citing strong APAC growth potentials and attractive tax schemes.
And since I left, Chinese talent, especially working at family offices and tech and AI firms, have migrated from Hong Kong and Shanghai/Shenzhen to the roaring lion city in droves. It felt like people spoke equally as much English as Mandarin in the streets.
It’s always shocking to think that all of this - meaning the making of the nation - only happened around 60 years ago – how the country came out of the British empire, briefly joined the Federation of Malaysia, and then chose to strike out alone. In that context, it makes more sense why some people still hold on to those markers of status and why there is confusion about identity, at least how I felt six years ago, especially among older generations, as my old ethnic Chinese Singaporean colleague proudly stated she was “raised British”.
But today, Singapore pride is much clearer among millennials and Gen Z, as what it means to be Singaporean has evolved into a multicultural, prosperous, and harmonious nation. And the epitome of all this is Keith Yap and his podcast.
At the airport, I picked up Kishore Mahbubani’s Living the Asian Century: An Undiplomatic Memoir. He’s probably Singapore’s best-known diplomat-turned-public-intellectual and a sometimes controversial champion of the ‘Asian century’ narrative. Putting aside policies, I was blown away by both his personal story, from the most humble of humble beginnings to becoming one of the most influential Asian diplomats.
What also really touched me is the way he explains his own success as intertwined with the timing of Singapore’s birth, much as many Chinese born in the 60s-70s identified their own humble success as tied to China's economic opening in the 80s-90s.
Instead of merely recognizing his own accomplishments, he told the story of Singapore’s economic planning and growth that enabled his ambitions and talent to be realized. For that, he compared his cousins who were born in neighboring countries that started out in similar impoverished circumstances, who never had the opportunities to achieve the kind of financial security and high quality of life he was able to attain through hard work. He came of age just as Singapore split from the British Empire, joined the Malaysian Federation, then separated again.
The early days of independence were marked by real poverty and existential uncertainty. What followed was intense state planning, aggressive education and talent schemes, and the very pragmatic choice to make English everyone’s first language to plug straight into global trade. The growth of the 60s and 70s set this tiny peninsula on a trajectory to become one of the richest economies in Asia, if not the world.
I guess there are a few themes I am touching on here, but probably will not unpack fully in this post: what it means to be a “citizen” of a country in a globalized economy; identity politics hijacking economic planning versus the crasser, more pragmatic reality of economic class differentiation and planning; state planning versus free market capitalism.
In many ways, Singapore has built something close to a technocratic utopia for the average citizen, even if that story looks very different for low-wage migrant workers or those at the margins. Singapore, at large is the epitome of the middle-class dream.
A majority ethnic Chinese population, plus Malay and Indian minorities, now broadly share a national identity after decades of tension and negotiation. The planning has sometimes been criticized as being artificial or surgical, but truly, it is crucial to take a step back and appreciate the social harmony the nation exemplifies. Especially amid the tensions we’re seeing in other major cities across the globe and now exacerbated by the increasing income gap.
I know my vantage point is partial. As social cohesion feels strong at a surface level, even though there are still real debates about race, class, and who gets fully included in the Singapore dream. But let us not forget Singapore has one of the highest concentrations of billionaires in the world. Yet the anger towards the rich actually feels much less prevalent than in cities like New York or Hong Kong.
Social cohesion feels strong because public housing is subsidized, spots at top universities are reserved primarily for locals, and jobs are protected through the Dependency Ratio Ceiling, which is meticulously calculated based on local labor capabilities for each sector. Young couples are incentivized to purchase homes and bear children as housing schemes favor first-time buyers. And not to down play it too much, but in essence, there is a housing supply shortage driving up prices in many major cities, which is making this generation feel so disgruntled.
Whereas Singapore’s government has essentially made sure more than 80% of all houses available are HDB housing, ensuring that local residents and citizens have the ability to own a home, whereas foreigners or higher-income people can choose to pay more (and be taxed) for fancier privately developed condos.
The state has created a path where each generation feels there is still hope: a decent job, a home, a reasonably good life - achieving the founding mission of Lee Kuan Yew. Affordability, which sits at the heart of so much discontent in the major cities of the U.S. and China right now, feels less like an open wound there. The tradeoffs that former Prime Minister Lee and his generation chose are now widely praised across very different political and economic camps. But sixty years ago, none of this was guaranteed. [actually came across this post that explains the HDB/ SG housing planning well, written by Varad Dongre *apologies I tagged the wrong person in the email blast*]
Today, Singapore is hailed as one of the safest cities in the world, but that safety comes from an extremely strict legal regime. The death penalty applies to offences that, in many Western contexts, would be treated far less severely – including certain drug offences like marijuana trafficking.
Even being seen naked through the window of your own home can, in theory, result in a fine. The line between order and overreach is thin and constantly negotiated, but the result is that the average citizen lives in safe and affordable homes with access to above-average education and access to affordable, clean food provided at hawker centers installed in every major neighborhood. Some may say it is only able to reach this kind of balance because of its population, which is only hovering over 6 million people. Maybe so?
Since becoming a mother, I am not sure if I have become more conservative or just more honest about my selfish priorities. I walk through any city now asking: Can my children be safe here? Will there be strangers approaching them, needles or drugs in the parks, random volatility on the streets? So when people ask why I choose to live in Asia, the answer has become almost disarmingly simple. I want a safe place for my kids to grow up. I accept that there are tradeoffs that come with that, including the sense of a more controlled environment and, at times, a feeling of stifled dissent or innovation. So every system has a trade-off, and maybe at different phases in life, we’re willing to give up different things for different priorities.
Six years ago I left Singapore, and to be honest, I didn’t really enjoy my time there back then. This trip felt completely different and ended up being fruitful on both personal and professional levels. I realized how much I’ve grown to appreciate Singapore’s efficiency – it literally took me seven minutes to get from the plane, through immigration, and out – and how much I needed a more nuanced understanding of its history and relationship to the British, China, Malaysia and the world.
For someone who spent the first two decades of her life predominantly in North America, this wasn’t obvious at all. At the time, both Canada and the U.S. were going through public remorse for past actions concerning mistreatment of Indigenous Peoples and minorities at the peak of the mid-2010s.
It shocked me to see celebrations of Raffles everywhere when I first landed in Singapore from New York, the contrast of two cities couldnt’t be more stark, where one was extremely silent, orderly and celebrating its colonial history (with a bicentennial exihibition of Raffles landing on display at Fort Canning), to a loud, and a let’s just say much less orderly city with a lot of people frustrated and angered at unfair treatments of racial minorities. But as I’ve spent more time in Asia, I’ve learned more about the identity struggles many cities and nations have had to face post WWII and then more recently, reinvent.
On a lighter note, I’ve also learned to appreciate its curated aesthetics compared to the more rugged and grand nature of the great white north that I was familiar with. Singapore is incredibly good at crafting/molding and artfully displaying nature: the Cloud Dome, Botanic Gardens, Gardens by the Bay, even the greenery in hotels all feel perfectly planned – almost artificial, but not quite, because the oxygen and the smell are still real. It’s nature with air-con, and without the mosquitos.
On my flight back, after many conversations and observations, I found myself thinking hard about the role of the state in the economy, how that shapes innovation and value creation, and how welfare, wellbeing, and subsidies interact with risk-taking. And how all of that actually is reflected in the aesthetics of the country - controlled, beautiful and practical.
People often say Singapore isn’t the place for innovation, that comfort provided by the state dulls the hunger to build. When there is too much comfort, people become content, and when they are content it’s hard to harbor true innovation. But the tech and startup scene felt genuinely vibrant this time. Alas, I believe Singapore will play an even bigger role as AI diffusion accelerates globally, serving as a key entry and exit point to the world’s fastest-growing region.
But of course, I recognize that this is an incomplete snapshot. I’m seeing Singapore through the lens of a visitor and a fairly privileged one. Anyway, that probably deserves a full essay of its own one day.
With that, I want to say thank you to all the readers who came out for drinks, grabbed coffee, or showed me around their offices. I want to thank the contacts at ByteDance, Google, Meta, and Tencent, as well as some more casual catch-ups with friends at CapGo.AI, Deep Intelligent Pharma, Manus, Final Round AI, and Liminal.
Now to the main takeaways from my meetings with people at Google, ByteDance, Tencent, and Meta…
Conversations with Tencent, Google, Meta, and ByteDance
This hypothesis I have about the next stage of AI diffusion emerged from a series of conversations with Tencent, Google, Meta, and ByteDance. The year of 2026 is when we’ll truly see diffusion being prioritized and it will be into different corners of the world in different formats.
Throwback to the 90s and early 2000s, the first wave of internet companies was basically catalogues. Yahoo, AOL, and portals that organized information for a tiny group of wired users. As infrastructure improved and became cheaper, the internet stopped being a catalogue and started reorganizing behaviour. Google rewired how we searched, Facebook rewired how we connected, Uber and DoorDash rewired how we moved and ate - changing the real economy - the physically touchable world.
The term “Internet company” eventually stretched from social networks to ride-hailing, logistics, payments, and food delivery.
It’s hard for me to believe some version of this won’t play out with AI. Right now, most people still equate AI with ChatGPT-style chatbots. These are extremely powerful for knowledge workers, but that is only a slice of the population (to be exact, about 1 billion globally).
As AI diffuses, the more interesting wave will be the one that seeps quietly into existing apps, tools, and hardware, and leaves space for very niche builders to pick up the pieces that big platforms don’t touch.
So coming out of these meetings, I started to see four main “rails” for how AI is diffusing, especially for big techs.
Tencent is doing it through social and emotional infrastructure, with WeChat as the nervous system. ByteDance is doing it through content and creativity, via short video, ads, and live commerce. Google is doing it through productivity and information infrastructure across search, mail, maps, and Android. Meta is doing it through hardware and embodiment — glasses and whatever comes after.
On top of that sits a bigger axis: in the U.S., AI is spreading in an enterprise-first way, while in China, it is more consumer / super-app-first. Once you overlay these together, the diffusion picture becomes much more interesting than “everyone ships a chatbot.” And for startups, there is still room for new breakthroughs, but it may lie within verticals - pharma, biotech, education, legal, and so on.
Let’s break it down.
Tencent: Looking beyond Knowledge Workers
The representative from Tencent said that in many ways, Tencent has always been slow to adopt new technology, but when they do, they do it very strategically. When TikTok/ Douyin took off with its short videos, Tencent watched how the game was played and launched WeChat Channels (videos) a few years later, and very quickly integrated the function within its e-commerce cycle. On that note, Tencent also watched Alibaba roll out its payment function Alipay and then only to strike later with WeChat Pay, solidifying its place as the second most used payment app in China, too. (I wrote about Tencent’s roots in fast-following here)
From their point of view, most consumer AI products today are still aimed at knowledge workers. ChatGPT clones, writing tools, coding copilots. (and I, myself, have been fixated on this too) Useful, but narrow. The majority of people in China are not “using AI” in any meaningful sense yet. Think about the 1.3-1.4 billion people on WeChat. How many are actually knowledge workers? Skilled workers make up about ~21%.
So she explained to me - look at lower-tier cities (outside glitzy Beijing, Shanghai, Shenzhen, even Hangzhou, Qingdao, Chongqing). First, think about people who do not sit in an office doing long, windy research and reports. Then think about the vast amount of (aging population) seniors who live there while their kids work in tier-1 or tier-2 cities. They orbit around different planets and stars, essentially. The parents feel both physically and digitally left behind. But almost all of them use WeChat. So inside Tencent, the question is: can WeChat use AI to rebuild some of these emotional and social bridges?
You can imagine an AI system that notices long stretches of silence in a family group, picks up on a grandparent’s lonely days, and nudges the child in Shanghai to call. “Your mom mentioned that recipe last week. want to do a quick video call and cook it together?” Or a companion bot that reads the news out loud, explains memes, or retells family stories in a familiar dialect. The idea is not to replace human connection, but to patch the gaps that are already there, inside an app seniors already know how to use.
Tencent’s advantage is its integrated ecosystem: messaging, payments, mini-programs, content, all inside one super-app. Once AI is stitched into that fabric, diffusion can move very fast. WeChat stops being “just a messaging app” and starts to look more like an emotional nervous system for people who have never written a single prompt. That’s a very different story from “one more chatbot product.”
ByteDance: Creators, Content, and Commerce
If Tencent’s rail is relationships, ByteDance’s rail is content. One thing that stood out is that ByteDance does not think of AI as a separate product line. It is treated more like fuel poured into a machine that is already running hard: short video, ads, live commerce, creator tools.
For creators, that means help with scripts, cuts, subtitles, translations, filters, thumbnails — basically compressing what a small studio used to do into a single person’s phone. For advertisers and merchants, it means generating lots of creative variations, localizing them, and testing them quickly across segments. For live commerce, it looks like agents that auto-clip streams into short videos, handle common questions in chat, and pass the complicated ones to human hosts or customer service.
The quote that was uneventful but also accurate is “it’s all AI, who isn’t doing AI, in many ways it’s all the same.” But what sets one different from another is the existing distribution and ecosystem. For ByteDance, that ecosystem lies within entertainment and creation, and more increasingly, ecommerce. The strategy is to push AI as an invisible co-pilot behind the scenes for creators, SMB merchants, and advertisers who already live inside that ecosystem all day.
Whether it is Douyin or TikTok, this is where people actually spend their time. That makes this a very different diffusion channel from “AI in productivity tools.”
If Tencent’s superpower is the social graph, ByteDance’s asset is the attention graph.
Google: AI as an Invisible Layer
Google’s view felt like a logical extension of this, but with a strong enterprise and productivity lens. The message was basically: Gemini is not meant to be just another ChatGPT competitor. (Although the recent Gemini3 has been wowza) The real question is how AI shows up inside every Google product people already touch every day.
So instead of building a new AI destination, the focus is on an invisible layer across the existing ones. One theme here we can identify is to trap all users within its existing ecosystem.
For Google, it’s about seamlessness (and trapping everyone’s ass). Search becomes less about keywords and more about understanding intent and context. Gmail drafts replies in your tone. Docs and Sheets quietly summarize, restructure, and link content without you asking. Maps nudges you away from storms or traffic. Android becomes a voice-first assistant for people who may never type a long prompt. It’s the seamless integration of AI, and the next step is proactive support.
For billions of Android users who are not sitting behind desks — people in factories, farms, and informal jobs. AI diffusion might look like voice-based crop advice or weather alerts in local dialects, simple routines that turn photos and voice notes into invoices or expense logs, or AR overlays on a phone camera that help diagnose what is wrong with a machine and which part not to touch. People don’t need to go learn a new thing called “AI”; their existing tools get more helpful.
[Thank you, Mani for the awesome tour!]
On top of this, Google is moving toward agents that can actually do things across products: booking trips, organizing emails, and collecting documents for a brief. It’s like when Google integrated hotel and flight booking features - it’s because people were searching on Google for these links already. Google knows, before we know what we need (creepy, but real). What Google can do is move away from just “ask a question, get an answer,” to “state the intent/ result and let the system handle the execution.”
Meta: Hardware-Software Integration
Meta brings in the hardware angle. Their bet is that AI will not stay trapped in screens and Meta AI will live in devices we wear. Beyond Meta AI functions for creators, the company is pushing hard on smart glasses, and I was invited to a closed-door soft launch of their Meta glasses in Singapore.
I got a pair of the RayBan x Meta glasses in SF last year, and they’re great for snapping pics of my toddler running around Disney, but I still feel very, very icky about the safety issue here. Yes, there are lights on them, but who’s sticking their face to yours to see if your lights are on?
Anyway… One example, one of the panelists- the Chief Product guy at Grab (SEA’s super app), talked about that stuck with me is food delivery workers. Right now, riders in dense cities juggle multiple apps on their phones, glance down at maps while riding, and constantly check orders and addresses. It is stressful and dangerous.
If those riders had AI-enabled glasses, their day would look very different. As we orient the physical world with our eyes, the smart glasses will be able to be an extension of your eyes - like supercharging your eyes.
Navigation can sit in their ears/ on their eyes balls and not on a screen. Subtle prompts tell them when to turn and when to watch out for a car behind them. On-device vision models can confirm order numbers and addresses in a split second. Signs and notes in another language can be read or translated in real time. Instead of “phone in one hand, bike in the other,” it becomes eyes-up, hands-free work with AI woven into the background.
The current Ray Ban x Meta glasses still feel a little nascent, but you can already see the direction. As hardware improves and models get better, glasses move from being a fun gadget to serious safety and productivity gear, first for gig workers and frontline staff, and eventually for more people. And glasses are only one piece of embodied AI. For Meta, the ecosystem within Facebook, Instagram, and WhatsApp - hardware may make sense.
It connects directly to EVs behaving more like robots on wheels, warehouse and delivery robots coordinated by AI agents, and home or service robots that handle care and basic tasks. Meta’s role is to show what happens when AI stops being something you “open” on a screen and becomes something that is simply present in what you wear and carry.
So were Snap’s glasses maybe just too early for the market? And Chinese wearables like Rokid glasses and the likes of iFlytek moving into simultaneous translation glasses probably have a huge upperhand in cost but again, their then gain, ecosystem is not as big and to get a user to download another app on their phone will be a challenge.
Enterprise vs. Consumer Rails
Looking at it holistically, U.S. big tech still focuses on the enterprise-first way. Think Microsoft, Google Cloud, AWS. AI is bundled into cloud, SaaS, office suites, and coding tools. The first big wins are dev productivity, office productivity, and internal agents. In China, AI is more consumer and super-app-first. Think Tencent, ByteDance, Meituan, Kuaishou. AI is baked into social, feeds, payments, logistics, and services. The early wins are engagement, stickiness, and monetization inside those ecosystems.
Underneath all of this is a quieter shift from chatbots to agents. The way I think about it is: chatbots were the UI of the dial-up era. You go to a site or an app, type something, and it responds. Great for research, writing, coding, and planning. Agents are closer to broadband. They sit in the background, watch context, and can act across apps and services, only interrupting you when they need a decision or consent.
In Tencent’s world, that might mean WeChat agents that handle bookings, payments, translations, summaries, and reminders across mini-programs. In Google’s world, it looks like Gemini-style agents that clear your inbox, pull the right documents for a meeting, draft the notes, and schedule follow-ups. For ByteDance, it is agents that automatically produce and distribute content, test creatives, and manage spend across campaigns for small merchants and creators. For Meta, it is agents that see what you see through cameras on glasses or phones and help you navigate and stay safe. The surface may still look like “chat” in many places, but under the hood the shift is from answering to doing.
Internet Pattern, AI Rerun
Echoing what I started the article with. Internet companies started with catalogues… and look where we are now with AI.
Zooming out, my mental model right now is that AI is still in the very early innings, and we are just starting to see the diffusion waves that rhyme with the internet’s rollout. In the early days, AOL and Yahoo defined the mental model: portals and directories. Then came email, search, and social. Today, when we say “internet company,” we include ride-hailing, e-commerce, logistics, payments, streaming, and a lot more. The internet ended up reorganizing the physical economy as much as the digital one.
AI seems to be on a similar arc. The first phase is the “catalogue” phase: ChatGPT-style bots and standalone assistants are used mainly by early adopters and knowledge workers. The next phase is more like “information dissemination,” where we saw the normalization of emails and the digitalization of media: AI quietly seeping into ecosystems like WeChat, Douyin/TikTok, Google’s productivity tools, and Meta’s hardware — becoming an invisible co-pilot instead of a separate destination.
[See illustrations generated by Gemini]
Beyond that is the “tangible reorg” phase, where AI starts to directly reorganize parts of the physical economy: glasses and robots for workers, elder-care fleets, AI-optimized farms, safer construction sites, vast-data supported medical diagnoses instead of relying on the memory or know-how of one practitioner’s brain.
The Gaps In Big Tech: Opportunities for Startups
Across Tencent, ByteDance, Google, and Meta, there is a common thread: all of them are focused on deepening value capture inside their existing ecosystems.
WeChat’s social and payment rails. ByteDance’s recommendation and content engine. Google’s productivity suite and cloud. Meta’s social graph and hardware. They will naturally prioritize the obvious integrations first, where they already have data, distribution, and revenue - it’s about “trapping” us in their ecosystems.
That still leaves a lot of gaps. There are users and communities that do not fit a one-size-fits-all UX. There are vertical workflows that are too small or too specific for hyperscalers to bother with right now. There are local languages, norms, and regulations that require boots-on-the-ground localization.
This is where I expect smaller players and startups to step in, as we’ve sprinkled throughout the piece. Areas like biotech, pharma, medicine, education, legal, and so on that require really precise industry knowledge and data will have companies to fill that up.
In the same way that Etsy, Shopify, and TaskRabbit emerged on top of the general “internet,” I think AI will produce hyper-specific agents and models for expat parents, street vendors, indigenous language education, and very niche professional workflows.
But most of these will still sit in big tech’s gravity wells. As in, using their models, app stores, cloud infrastructure, and payment rails, so you get room to innovate at the edges, but not complete independence. Even so, that’s where much of the upside and differentiation will lie.
Just read through your in depth insights regarding the few big tech players’ niche plays vying for different vantage points to dominate within the AI Universe. The many smaller edge players, though 100% dependent on the big ones, are indeed indispensable boots-on-the-ground symbiotic partners, who can intimately take the average everyday consumers (end user like me) by hands, step by step. At the end of the day, the devil is in the details—reliable and consistent data. Trustworthy and fully integrated data (structured and unstructured), along with general common sense ethics, are the secret sauce for achieving sustainable success within the ultimate human-led autonomous Agentic AI ecosystem.
Great reflections and connecting the dots as always, Grace! Hadn't expected such an energized view of SG innovation scene