Personal Report, 2025
Personal Report, 2025
Every year since I turned 16, I would take the final week of the year to read back my journals and logs, and write up an annual personal report to reflect upon the year. Writing is how I try to make sense of my life, so that it is not just fragments of experiences scattered across time, but one stream of meaning.
2025 was my first year building Artificial Societies. We went through the YC hype, the seed-round buzz, the post-batch slump, the endless reasonings and pivots, and now - touch wood - a simple clarity of focus on making something exceptional. It felt like a decade had gone by in a flash, and I found myself suddenly living in a different world, where I am a different person.
This is my reflection on 2025, this year of learning: how I learned to build Artificial Societies, learned to empower and be empowered by my wonderful team, and learned to live beyond my business, as my own expansive self.
Patrick, Tom, and I started the year in SF for YC. At the W25 batch kickoff, I told Garry Tan that I felt a sense of excitement I had not felt since before Cambridge. I found the future excitingly uncertain, for in this uncertainty, there were infinite possibilities for our own making.
Indeed, the future of Artificial Societies was entirely uncertain. We had no customers, no revenue, no clue how we’ll provide value to real people. Yes, we were the kind of startups they warned investors about: a hammer-on-paper, looking for a nail.
We spent our time at YC building demo products to ‘Make Something People Want’: first, a simulation of a thousand VC investors for founders to simulate how investors react before pitching; then ‘Reach’, a tool that simulated LinkedIn networks, so that authors can test how their posts land before hitting publish.
Reach attracted over 10,000 signups in days, gave us our first income, and showed enough promise that got us great seed investors. Because of Reach, we had a practically infinite capital runway, a database of over 1 million AI personas based on real human behaviour, and a LinkedIn brand that was getting millions of impressions - after all, we had a product built for it.
I’m proud of what we achieved with Reach. But when we returned to London in April, we realised that it wasn’t working, and that we needed to experiment more. We decided to take a big bet - build a unified social simulation platform, a ‘horizontal product’ with a range of ‘vertical applications’, so that we could let the world show us where to focus.
We launched the brand new Artificial Societies in late July: you could simulate any group of professionals, and see how they’d react to nearly twenty different types of content. Our launch got millions of impressions, just as we had simulated. But in the following weeks, neither trial conversion nor churn was looking good, and when we spoke to our users, everyone said something different.
Turns out, in the hope of proving that simulation is “Something People Want”, we built a product full of promises, where some pulled us away from our mission. Some users wanted an instant answer of whether content A or content B did better; others even told us to just give them the best possible content on autopilot in their writing styles, and forget about social network simulations altogether.
We made another mistake: in response to user feedback, we cut functionalities and built new technologies for each application, some entirely unrelated to simulations. We were spread too thin trying to maintain all these entirely different technologies, and struggled to actually deliver true value. Here’s the challenge with building a horizontal tool: sure, you could use a Rolex to smash walnuts, but the walnuts can push you to ruin the Rolex.
Around late September, we knew that we needed to focus on one ICP and one application, but we were lost in the maze. At times, it even felt like I wasn’t building Artificial Societies anymore, but someone else’s startup that I had inherited, trapped in someone else’s dream. After nearly a month in the ‘Pit of Reasoning’ and building different prototypes, we arrived at a new plan. We wrote all our thought process down in a big decision doc, ready to get back into build mode.
But the universe had other plans for us.
In late October, a major enterprise project came in. They wanted to use our simulations to help a large US company model how policymakers and opinion leaders react to different AI strategy narratives.
This was something we had ruled out in our reasoning for reasons: “we are bad at sales”, “enterprises would always trust traditional surveys over simulated surveys”, “simulations are best for tasks where speed matters”… But this project defied all our reasoning: a major enterprise trusted simulated surveys enough to want to base a major decision on it, and they didn’t mind if the simulation took minutes or days, as long as it gave them comprehensive insight on how their stakeholders might react.
We realised that this is what Artificial Societies was meant for: high-stakes research where all the traditional alternatives are prohibitively slow, expensive, or simply too risky. Because of Artificial Societies, impossible research methods - like testing how policymakers would react - are now within reach.
The whole team jumped on the project. With my former flatmate and co-author Felix joining us, we made more research breakthroughs in November than in the entire first year of Artificial Societies. Instead of removing functionalities to get results in seconds, we built more intensive simulations that took hours, but made use of our Behavioural Science background to capture more nuanced psychologies - everything from collective emotions to qualitative insights from every single individual.
Another similar project came through. And another. Even though our landing page was still the old product, even though we’ve done 0 outbounds and spent $0 on paid marketing, and even though we kept increasing our pricing, our pipeline started filling up in December, to the point now where we are running out of capacity.
It’s funny how all our planning and reasoning went out of the window the moment reality hit. Instead of a horizontal product that required many technical solutions, we now have one product that does the right job for our technology - simulate survey responses of high-value professionals - insanely well.
Looking back, we realised that we needed to focus on exceptional problems. Sure, our technology might be able to solve many problems, but only a few of them can push us to build a more exceptional product and technology, attract more exceptional partners to work with us, and drive us to become a more exceptional team.
When 2025 began, I was on my own in SF, shipping my janky code into production. Tom had just joined us, and was now going to move to SF to live with myself and Patrick for three months.
I had never led a team before founding Artificial Societies. Those first few weeks of W25 felt like I was pretending rather than actually being a leader - I’d try to go through the motions of leading meetings, being careful not to be too pushy towards the team, while also feeling like I should ‘manage’ them.
One day, about a month into the batch, it suddenly occurred to me that this was not just my company; it was Patrick’s and Tom’s company, as much as it was mine, and it’s actually quite nice that I wasn’t just on my own in my life’s mission anymore, like I had always been since I left home at 14 to go overseas.
But it took much longer for this realisation to properly set in.
During the Demo Day fundraise, I tried to take on everything myself to protect the team from all the turbulence. I got very stressed, needing to reply to all the investors, do the same pitch over and over again, and at the same time, do my engineering duties and fix bugs in production.
Over dinner, I told my friend and fellow founder Brian how stressed I felt. He stopped me right there: “You can’t protect the team from fundraising. It’s something that matters to the business, a business that they are personally invested in. You can’t fundraise or build this business alone; you have to be accountable, and involve your team to help.”
It was a godsend. Following Brian’s advice, I got Patrick to help reply to investors, and got Tom to help format the deck. We became oversubscribed within a week, three days before my 24th birthday. Tom and Patrick got me a huge birthday cake - which they then happily consumed when I found it too sweet.
Still, I felt that as CEO, I needed to have clarity at all times. When we returned to London and felt lost, I made a big plan of how to shape the product, and mastered the confidence to push it onto Patrick and Tom - I was ‘the CEO’ after all, I needed to take bets.
But in reality, I secretly wavered over every decision, fearing that I was making so many mistakes and leading the team in the wrong direction. It all came down during a team coaching session: Patrick and Tom weren’t really convinced of the new direction, and I almost had an emotional breakdown.
I couldn’t thank our wonderful coach Katie enough for how much she taught me about leadership. She showed us that each one of us was an aspect of Artificial Societies, and our moods reflected the business’s needs - it wasn’t just James who wanted a clear direction, it was a need of the business that would manifest through someone else, even if it wasn’t James.
Over the following months, she slowly taught me how to lead through communication; how to communicate what I needed psychologically, so that I work better with the team; and how to understand what really mattered to the team, even when they did something I disagreed with. She helped me reframe my fears and insecurities into awareness and power.
Slowly, instead of hiding behind a shell of ‘leadership’, I learned to be vulnerable to the team, trusting that they’ve got my back; instead of pushing a decision onto them, I learned to let conversations happen, and let the right decision take shape under my ownership.
This idea that everyone is an owner of Artificial Societies grew into our team philosophy: instead of having the founders make every decision, and passing directives down a hierarchy of specialists for execution, we believe in empowering a small team of exceptional individuals to go beyond given ‘roles’. Who says engineers can’t do marketing, and sales can’t ship code? We want to build a team of truly exceptional individuals, empower them to defy expectations, and live their life’s mission through Artificial Societies.
Over these months, though we struggled through great uncertainties, the team bonded in the trust that we are all owners of Artificial Societies. I noticed that, unlike the overbearing anxiety and wavering decisions I had at the start of the year, as the team took on more responsibility, I also became more confident, filled with optimism that no matter how long this would take, we were going to make it, because we were an exceptional team.
At the end-of-year peer perspectives, the team told me how they feel more confident about our future when “every morning James would bring his passion and conviction for AS”. I am grateful to have come such a long way since I anxiously repeated that I must “have the courage to make mistakes” - turns out, a team bonded through trust and shared conviction, can face anything without fear.
Since leaving home at 14, there were only two times when I worked as much as when I was at middle school in China: the first time was my final year at Cambridge, and the second time was this year of building Artificial Societies.
At first, I was driven by a pure fire of excitement. Sure, it was stressful, but the opposite of stress wasn’t happiness, but boredom; and no matter how stressful building a startup is, at least I wasn’t bored. This was my life’s work after all - all my interests finally unified into one - it’s incredible that I get to do this every day, and for the first time, I felt like I could work all day and all night, because my whole self is within it.
But building a business is not an easy task. Since I identified myself almost entirely with the business, any setbacks, any struggles, became personal setbacks and struggles. For months, I’d have massive mood swings depending on how I perceive we are doing, sometimes dragging the team down with me.
A couple of weeks before W25 Demo Day, we walked through Golden Gate Park and went to the beach. The three of us sat on the sea wall, over the yellow sandy beach, and gazed towards the expanse of the Pacific. Afar, surfers swam out to catch waves. The great waves, made by the moon’s gravity, travelled thousands of miles with the wind, to meet us here. They sounded like the ocean’s heartbeat, and I felt calm listening to its rhythm, as the sun warmed my face and drowned my mind with afternoon drowsiness. I felt a sense that there’s so much life out there, so much life to experience and to celebrate.
Those days at the end of YC W25 were glorious. But as the batch ended, my mood took a turn. I didn’t realise how much I got used to living with Patrick and Tom - my tribe - and how much I needed a personal life outside of Artificial Societies. Getting back into building helped, though somehow, I still refused myself a personal life.
It took Tom and Patrick a few months to help me see that to give our 100% to Artificial Societies, recovery is just as important as working hard. I slowly started meeting up with friends at the wonderful Culpeper pub - where I later bumped into Jensen Huang when he came for a Sunday Roast. And now, every morning, I play some music on my Guqin, currently arranging Hotel California on this ancient Chinese instrument.
I took my first proper holiday in November and learned to surf in the Canaries. Once, I swam out, deciding to make my best effort to catch a bigger wave behind me. I paddled hard to not miss my shot, and the wave took me up. It was faster than I had expected - the nose of my surfboard dived into the water, and I was swallowed into the wave, spinning and spinning. I counted to stay calm, and reached towards the light to find my surfboard. As I emerged and took in a deep breath of the fresh, sea-salted air, I laughed out with joy: how wonderful this life is, this life of so many experiences.
At that moment, I suddenly remembered the time when we went for a cold plunge in SF. It was after W25 Demo Day, and we went to the beach with our YC batch mates. When I stepped into the cold water, my whole body tensed up. I breathed heavily, feeling every cell in my body screaming at me to get out of the water. But in a moment, that feeling subsided, and I found myself calm in this gentle wave, watching the sunset coming through Golden Gate Bridge, landing on the blue water and becoming a thousand dancing belts of light. I noticed my awareness of how cold it still was, but I could separate my mind from the cold, remain aware of the external temperature, yet free from its control.
I realised that building a startup, as a way of life, is like being at sea. There will be ups and downs, and you could let mood swings take control, or you can remain aware but in control of yourself. After all, that’s why I am building a startup instead of having a “chill job” and a “chill life” - I want to be the one who defines my life’s constraints; I want to be the author of my own destiny; I want to pour my soul into making something exceptional, and let it go out there to take part in this beautiful world of humans.
Slowly, I stopped needing to practice mastering conviction, and started naturally living and breathing this great mission of building Artificial Societies. I see every moment in our product, and our technology, something to be curated with our full heart, our gut instincts - so that we train our full self to live with intentionality, and build a product with personality, a product that maximises our participation in this world.
Looking back, I realised that I have finally grown from being an engineer to being a CEO, living and breathing in the conviction and driving the team forward. And thanks to my wonderful team, I can now have a life beyond the business, and let the business grow beyond me.
In our first-ever YC Office Hour, Jared, our partner, told us that we should focus on figuring out how to deliver value first, instead of going straight into driving revenue. It’s funny that it took us a whole year to let it sink in.
Of course, I still have a lot to work on. I need to get better at setting mid-term goals and strategies, and communicating my thoughts to the team more proactively, so that the team has greater clarity on our future. It’s something that was there even at the first peer perspective of 2025 - something I still need to work on. But every day I wake up feeling so grateful that life took me here, where I am working with an exceptional team to work on exceptional problems, and build an exceptional product and technology.
Now, on the eve of 2026, we are finally beginning to deliver real value to real businesses: providing comprehensive awareness on how their most important stakeholders would react to their most important decisions. With the growing list of enterprise partners, our January is already looking like an insane month - and who knows what the rest of 2026 will hold? The best we can hope for, is that whatever our 2026 turns out to be, it will defy all our expectations.
December 31st, 2025
London