Jon Fosse, a Norwegian author and playwright, won the Nobel Prize in Literature in 2023. I have limited experience with Nordic culture, but Jack and I did travel to Iceland in 2020, and one of the best ways to understand a culture is to read stories, so I picked up Jon Fosse’s most critically acclaimed work called Septology (2019). It’s a 7-part novel spanning 3 volumes, exploring an aging Norwegian painter’s life through his doppelganger.

It took me over a year to read. Honestly, some parts really felt like sludging through icy mud; quite fitting for the setting of the book. It got better towards the end, and was more rewarding than I initially thought it would be. It had very religious themes, it was depressing, a bit schizophrenic, and it was written as one long sentence (without much punctuation periods), and at the end it made me reflect on some philosophical and existential parts of being alive.

The writing style, written as one really long sentence, made it seem like it was a breathless prayer to God. The main character is called Asle, but you also meet another character named Asle later, which is what made it feel schizophrenic. Reviews say they are doppelgangers, both painters, with one who quits drinking and starts to exhibit, while the other is drinking himself to death. Personally, I felt it was more like mirror worlds, one built by God and one built by the Devil.

I am in a different season of life these days, compared to my past. I envision life like soundwaves, and you either live life where the highs are very high and the lows are very low, or it’s oscillating very close to flatline. My life these days is much closer to the latter. I can’t say definitively which one is better or worse, because you probably need both in different seasons of life so you can truly experience the full range of what living can offer, and while I generally think it’s healthier and more sustainable to live a more stable life, sometimes I don’t feel as much growth.

The book was great for not being performative or trying to be easy to read. It seems like he wrote it more for himself than for an audience. To me, creating art and being an artist is about fully being yourself and being able to express yourself in a realistic way. One thing I didn’t really like was how all the female characters were orbiting around the male characters, rather than having their own complexities. I think I would give this book 3.5 stars out of 5.

Circular rhythm of life, one long breath, pray you’re on God’s side.

She was less anxious or stressed than she imagined, and she felt calm. Prepared. She knew what she would talk about, and the story of her life was robust enough that there probably wouldn’t be too much back-and-forth during the first chat with her new therapist.

In the waiting room, Jocelyn had her hands neatly folded in her lap, blissfully ignoring the other patients around her, and daydreaming about being perceived as a better person after going to therapy. It would give her more social points, given the current political and social climate.

The receptionist brought over a few sheets of paper with a plastic backboard to write on. “Since you’re a new patient, you’ll need to fill this out thoroughly,” she said, avoiding eye contact and shuffling back to behind her desk. Jocelyn didn’t mind her mousiness. She murmured a thank you and started scribbling through the packet. She wanted to stay quiet and discreet.

Jocelyn had come in during the weekend, wearing her weekday brown loafers and a tailored blazer, knowing most clinics were quite cold in Taipei. She recycled a lot of her workweek outfits for the weekend as well, preparing for her eventual role as an executive when the company felt she was ready. Anyway, she wanted to focus the conversation on work-related matters, because there really wasn’t enough to chat about outside of it.

She stood up and brought the form and board back to the reception desk. She leaned in to the receptionist with a lowered voice and said, “Excuse me, I’ve filled out the form. How do I pay?” The receptionist pointed to a sheet of paper taped onto the desk that instructed patients to pay with a bank wire transfer. Jocelyn feel a stab of annoyance.

Sitting behind Jocelyn was an elderly woman, mumbling to someone who looked like her son. Suddenly, she started raising her voice, demanding to see her doctor, and accusing the clinic of holding back her appointment to punish her. Other people in the waiting room shifted in their seats. Jocelyn watched with a careful eye, making sure she could reach an exit if the situation escalated. She also felt sorry for the mother and son, knowing that they needed the support here more than she did. Why was she here again?

Return to the character for whom you filled out the questionnaire. You’re going to put this (now fictional) person into the world and let him reveal himself. Imagine this person is entering the waiting room of a therapist’s office for the first time. The type of therapy is up to you (it may even involve bringing a spouse or a pet), but chances are this person will be feeling a little stressed. Keeping the character in the waiting room, write a passage where he is revealed through all four of these “showing” methods—action, speech, appearance, and thought. For a bonus round, you can put this character in an even more stressful situation, like observing someone being held at gunpoint. What will your character do then?

For the product meetup, I worked with David Tseng to help organize it. He’s started a few businesses and is mainly focused on the tech and product side. He was at the marketing meetup, so we were discussing what we could adjust to fit the theme and audience better for a product meetup.

We had a good deal of registrations and it’s unfortunate when we can’t accept everyone who wants to come and learn. Due to the venue and also the number of mods we could get, we prioritized those with product experience so that people can share as peers. Despite the rain, almost everyone came, which was great.

We had sign-ups from people working at TSMC, NVIDIA, ASUS, Mastercard, Coupang, Klook, momo, 台灣大哥大, 信義房屋, Synology, 17live, Appier, ShopBack, Gogolook, XRSPACE, Gamania. 60% of the group were product-related or founders, and then it was a mix of marketing / business, engineering, and designers. Interestingly, there was also a licensed lawyer in the audience.

I’m also very appreciative of John Fan who let us host our Claude meetup at their PicCollage office, which is a very nice space, and there were group chats in the fun meeting rooms, like the KTV room, or the gaming room, etc. and I thought it was very cute. In David’s intro, he mentioned that this kind of event format is great for introverts to be able to have the chance to speak at least once, in small groups with like-minded individuals.

We did two rounds like last time, but this time, David pre-prepared the groupings, assigning everyone to their tables and mods. He shared the groupings on the projector so people could find which table they needed to be at for each round. Afterward the two rounds, some of the audience members, including moderators, went on stage for short lightning rounds to share some insights.

One of them shared about how they had a personal obsession with baseball and knew everyone in Taiwan was really focused on baseball last year. He created an app where he was able to get hundreds of free users, a good number of paid ones, and made some money off of it. Now he wants to know if he can turn this into a company.

Another audience member shared about a concrete, replicable takeaway. Ask AI to generate role-specific questions that you answer yourself and then synthesize, because AI still can’t do a lot actually. For example, if you encounter a problem, instead of asking AI how to solve the problem, ask AI to generate a bunch of questions for yourself to think of the answers, and then consolidate your answers in AI to produce a summary and an idea of the solution and direction. I really liked this one.

There were also mods who talked about GTM ideas like living off your land, using B2B tactics for B2C and vice versa, and milestones as cutoffs. Knowledge-into-skills. Another mod recently broke his arm but was a good sport and came anyway. He had been testing out a surfing-related app he had created, by going surfing, but there was a storm and the winds were too strong, which is how he broke his arm.

Hosting this made me realize that people have very different styles of hosting events. David prepared a lot, with one of the mods even drawing the layout of the venue on Excel, a few days before the event, for the other mods to have the lay of the land before gathering there.

Related to product: I’m also listening to this podcast by Lenny’s Newsletter with Peter Deng who was the product lead at a number of iconic products: ChatGPT, Airtable, Uber, Oculus, Facebook, Instagram, Google. He has some great insights into hiring and building great teams. Will share more in a future piece when I’ve had more time to gather my thoughts.

Thanks to David, his partner Yumo who was great energy and supported a lot, the mods that David was connected to, John Fan for letting us use PicCollage, the attendees for being involved, and the Anthropic team for supporting these community events.

The next Claude meetup will focus on 3 founders who will demo about using Claude: Tammy from Numbers Protocol, Howie from Protico, and Eric from Apersona.

You can follow along at my WhatsApp channel for future updates: https://whatsapp.com/channel/0029VbDF8lb1HspurCJNEN0F

Unsupervised play is when kids play without adults watching, managing, or trying to intervene. It’s a little related to unstructured play, which is when adults don’t come up with the activity itself, but unsupervised play is when adults aren’t involved at all. Of course, in the state of the world today, and our hypervigilance and hyperawareness, this is not that easy for parents, including myself, but I still want to talk about the importance of it, and how we could potentially build spaces that allow for controlled, unsupervised play.

For those similar to my age, in our 30s or older, we can probably remember many times when we were kids when we had unsupervised time. For example, when I lived in Egypt, I would be able to walk from school to the nearby bookstore to buy my books and stay there to read. Living in Taiwan is also a privilege in terms of safety, and when I moved here as a teenager, I was able to do a lot outside by myself or with friends, like commuting over an hour by taking the bus and metro from school to home, or going to the movies, or getting food and drinks, by myself or with friends.

Lyor is only a little over 2 years old, so I haven’t yet grappled with how early I want him to start experiencing unsupervised play outside and independence as an individual person. There have been many moments when he’s playing by himself, whether at home or outdoors, when I’ve unintentionally intervened saying be careful before anything happened, but I try to catch myself before saying something these days.

A pair of Norwegian researchers found that kids naturally fear the things they’re not ready for, and risky play is how they dismantle those fears themselves, a little at a time, on their own schedule.1 When adults deprive kids of the chance to be scared, the fear doesn’t get resolved and doesn’t disappear over time. At the same time, there is a lot of research showing how free play has collapsed over the last 50 years, at the same time as depression and anxiety in children have increased to an all-time high.2 Correlation does not mean causation of course, but there could be some relation between kids’ independence and their happiness in life.

In Taiwan, maybe under the traditional idea of 「勤有功，嬉無益」, children’s right to play is often dismissed. The 兒福聯盟 (Child Welfare League Foundation) surveyed Taiwanese kids and found 33% of children get less than one hour of play on weekdays, rising to 52.7% for kids aged 9 to 12, as school hours, homework, and 才藝補習 squeeze out play time. 59.8% of parents believe play should have an “educational” effect, and 52.5% believe they should arrange activities so their child is never idle.3 I remember growing up, pre-screens and pre-Internet, how often I would get bored, either by myself or with my family and friends, and we’d have to come up with creative or curious ways to spend our time. I think this was a net benefit for all of us.

Being overprotective is very natural for parents, so I have to constantly be battling my instincts to help Lyor and find things to do for him to do, while being aware of the facts of what overprotection will do for his long-term growth. Also, sometimes we just subscribe to a system and we forget that. A system with real fears, where families are smaller and there’s less “sharing” amongst adults to support other people’s children, and mostly just two working parents who may lack energy to think more about these things.

Still, I believe it’s something worth bringing up and looking at. Taiwan’s future isn’t just about the much-debated fertility rate, but also a matter of how to raise children in the current environment we’re in, and a true lack of spaces that allow children to experience life on their own with controlled safety measures. Spain has these streets filled with restaurants with outdoor seating, and in between on the streets, where cars can’t access, are playgrounds so parents and friends can have a bite and a drink while being able to observe their kids from afar.

I do love some of the kid-friendly spaces we already have in Taiwan, like the outdoor parks and indoor playgrounds scattered around Taipei city, but I’m still looking forward to more family-friendly spaces that are good for the parents to do their own thing too, while allowing kids the freedom to play on their own, without structure and without constant supervision.

Ok, so that was a lot of AI-related jargon! It’s a little hard to describe to those outside the industry, and personally, I’m still learning about it, so I’ll get better at talking about it in the future. For now though, that’s a one-liner to describe what the company does. For me, I mainly focus on marketing initiatives here, but also lead the Fellowship Program investing in young talents in Taiwan, and am involved in many other aspects including product, BD, and hiring.

It’s been a while since I did full-time work, and I wanted to give a shout out to Guillaume from Headline who introduced me to this opportunity. It’s been really tough, to be honest. It’s a Silicon Valley style startup (we also have offices in Palo Alto and Singapore), and since it’s an early-stage company (less than 2 years old), the work and culture is intense af. Balancing that with family time has been hard, and I’ve given up a lot of personal time and hobbies to do this role well. I often work 12-hour days, and do work on the weekends as well.

However, after working at AppWorks and seeing so many founders and learning how to invest in great teams, I feel super bullish on Zettabyte. It can feel chaotic at times, and the product is ever-evolving, but the team is incredible. I won’t mention people by name, because they are very private people, but one guy who is younger than me exited 3 startups before the age of 30. Another studied comp sci at Berkeley back in the 90s then dropped out after selling his search optimization company for a few million dollars, and then went to law school and became a highly prestigious lawyer in the tech scene. One of our tech leads did his PhD in physics at Berkeley when he was 17 years old, another engineer was an MD surgeon before she decided to get into computer science, and a technical ops guy was building autonomous vehicles back in the 90s.

The team is stacked. That’s one big reason why I want to keep working here, even though these past 3 months felt more like 3 seasons. I get to learn from these incredible builders and thinkers. It’s not just the leadership team though, because my coworkers are such a pleasure to work with. I got to pull in Eric and Hazel, who have helped so much on the product and customer support side.

Another big reason I feel strongly about Zettabyte is that they are very mission-driven. So much of the global AI infrastructure currently relies on a few countries — what about everyone else? We want to support anyone who cares about owning and managing their own data and AI needs. It’s about the freedom of choice. It’s one of the reasons why we’re building in Taiwan as well. With Taiwan’s hardware capabilities, we can help support the software layer to power the AI ecosystem here and ultimately power the ecosystems everywhere else. We’re working with several local research institutions, thinktanks, universities, partners, communities, enterprises, and startups so we can together set a standard for what software companies can look like and do for Taiwan in our AI future.

I still want to keep hosting Claude meetups, and I have something planned every 3 weeks until the end of the year, to give back to the community that helped me find a role at Zettabyte. I apologize in advance for not being able to meet up with friends regularly anymore, since my time is extremely limited these days balancing work and family. And for those who use a lot of AI and need cost-efficient compute or GPUs, or want to recommend any great talents for full-time or the Fellowship Program, don’t hesitate to reach out. Come visit our office in Neihu and check out our GPUs yourself :)

When I moved to Taiwan around 14 years old, I was able to read and watch a lot more collections — usually illegally downloaded because I needed them in English, and Taiwan only had Japanese and Chinese versions. I got really into fanfiction at some point too, even writing my own once (it was terrible lol).

Anyway, over my lifetime, I was able to finish some mainstream ones that I really enjoyed, like Slam Dunk, Nana, and Death Note. I watched cutesy ones too, like Cardcaptor Sakura and Mirumo de Pon. I started a few that I never finished, even though I really liked them, like One Piece, because it was too much of a commitment. Some I tried and couldn’t really get into, like Bleach.

In recent years, I’ve tried watching the more modern anime, like Attack on Titan (too dark), Demon Slayer (I liked the first season), and The Apothecary Diaries, an absolute banger of a show.

For those that haven’t seen The Apothecary Diaries, it’s a show set in a fictional country that’s very clearly modeled after imperial China, specifically the Tang dynasty. It centers around a girl called Maomao who originally worked in the pleasure district outside the palace grounds as a toxicologist, or a medicine specialist, and then is forced to work in the palace to help test for poisons for the emperor and his consorts. The other main protagonist of the show that Maomao meets in the palace grounds is a beautiful eunuch that’s very popular with the ladies, called Jinshi. He also happens to be high-ranking, managing the rear palace that houses the consorts.

One of my favorite parts of this show is that, despite many of the women characters being consorts or prostitutes or servants, they are all presented in a mostly positive light, while some other women, especially the villainous ones, have very complex characters or backstories. It made me very curious about the writer of this show, because sometimes in manga and anime, the female “protagonists” were more like secondary characters.

It was a delight to find out that the show is based off a light novel series by a woman called Natsu Hyuuga. Even the name itself is very woman-coded. It tracks that a woman writer would be able to and want to create such multidimensional female characters. It also makes sense that some shows with great female protagonists, like Inuyasha and Sailor Moon, are also written by women. A reminder to myself that I should try Fullmetal Alchemist, supposedly a great series, also written by a woman.

I thoroughly enjoyed The Apothecary Diaries. It’s both funny and deep, light and dark, with suspense and romance and cleverness in its storytelling. Each episode is short, about 20 mins, which made it easy to binge, but I would use a lot of self-restraint so I could savor the show longer. Each episode is about a certain mystery that Maomao tries to solve, and sometimes a case would span multiple episodes.

I was so happy when the second season came out. The first half of the third season will be coming out this year in October, with a movie slated for the end of the year, and the second half of Season 3 coming out next April. Really looking forward to it, and I highly recommend it!

I had always been a bit of a teacher’s pet. I really enjoyed school and the education system I was exposed to growing up (until I moved to Taiwan when I was 14 years old), and because the Middle East is closer to Europe than the US, I grew up under the British curriculum. At the British International School of Cairo (BISC), I did a lot of school and class-related activities in my free time. Basically, I was a total nerd, but thankfully, at a place like BISC, and because of my classmates in Egypt, I didn’t feel like it was a bad thing at all.

However, just because I was really into school didn’t mean I was particularly strong in every subject. Being the only Asian in my entire grade of ~50 students, I was, maybe somewhat unfairly to my classmates, perceived to be quite good at math.

Now, don’t get me wrong, I wasn’t bad at math at all. In fact, for a while, I really was very good. Somewhere along the way though, I realized my true love was English and the arts, and since I never had a math tutor or mentor prior to Taiwan, I didn’t get much better at it over time. But because I was good for a while, and I was known by everyone as a school-lover, I feel like that’s why I got selected as one of 4 students to represent BISC for a global math competition called ECIS, held in Belgium that year.

I had imposter syndrome then, thinking I wasn’t going to perform well at the different activities during the competition. However, the few months prior to Belgium, every team had to prepare a physical game that other teams could play when we got there. That was one of the activities where we could score points, and since this was about creativity, I knew I could help us win this one.

I spent all my free time designing and creating this board game. It was really big, around the size of a desk, and it was set on a wooden board that was designed like a suitcase so it could fold up and be locked. When you open up the wooden suitcase, it would lie flat on the ground. We used papier mache to create rolling hills and painted train tracks snaking through the hills and board. At certain regular intervals on the track, we painted different colored jewels throughout, with each color representing a different difficulty of a math question. As the players roll the dice and land on a jewel, they had to pick the respective jeweled card that had a math question and if they answered correctly, they collected a jewel of that color that we hand-made as well. At the end of the game, you counted up the points of the jewels and the highest score won. (At the time, I was really into the board game Trivial Pursuit, and my mom had just gifted me and my sisters teddy bears with matching horoscope jewels — mine was sapphire — so I used these two ideas to create the board game.)

Anyway, I was super proud of this board game. When we went to Belgium, I didn’t do so well in the other activities, and the games activity was the last one everyone played together. Some teams just brought a piece of paper and drew a hangman version for math questions, so I was convinced we would win.

Unfortunately, the judges thought we didn’t create the game ourselves, and since there was no social media back then and we didn’t have any recordings of the process, we were disqualified. Our hard work! Disqualified! Because it was too good! My faith in math was gone forever from that point onwards (or at least, that’s how I coped with not being very good at it moving forward.)

I wish I didn’t give up on math so early in my life. It’s such an important skill to be good at. There have been numerous studies done showing that math does shape individual success, and a lot more than people admit. One study showed that good math abilities at 13 years old led to more tenure, patents, executive roles, and doctorates. Even within high math ability, the top 1%, the top quartile outperformed the bottom quartile. Another study showed how ~45% of billionaires were in the top 1% of cognitive ability, concentrated in finance and tech. Of course, much of this is tangled up with elite-school access and money, but indeed being able to afford private math tutors at a young age is a privilege, not just in money but also in the foresight.

In San Francisco, the current CEO of YC, Garry Tan, has been fighting to bring algebra back to the city's middle schools. The school district removed it back in 2014 with supposedly good intentions: they wanted to fix the unfairness of kids getting sorted into higher and lower math tracks so early. But instead of giving more support to those who needed it, they just took the math away from everyone, and it didn't pay off. The district finally agreed to Garry’s requests, but the rollout has been so uneven that it's now creating its own patchwork inequity.

To me, being good at math can be a systemic generational wealth accumulation, a kind of non-biological evolution. A built ability, gated access, partly inherited, compounding across generations. That works out quite nicely for Taiwan, which is top 3 countries globally at math performance. It was unfortunate for me though, moving from another country to Taiwan and not being able to catch up. I hope my son Lyor ends up loving math, and at this early stage of being 2 years old, I think I see some positive signs. It’s one of the reasons we definitively want to spend part of his childhood here in Taiwan, so he can pick it up while young (alongside his Mandarin-speaking and Chinese reading and writing abilities.)

Anyway, apparently math can be picked up as an adult if we missed the window as a kid. The hard thing is less to do with the mental abilities but more so because we simply have too many other responsibilities to be able to carve out enough time to practice, which is important for math, actually more important than cramming. So maybe during some point in my life, when I don’t need to work full-time, maybe it’s worth my while to see if I can pick it back up again.

Her hair is black and her eyes are deep brown.

On her face, she has soft, round cheeks, a mousy nose, and a petite, smiling mouth. She has a large, flat, latte-colored mole on her right thigh. Her skin was otherwise unblemished; decades of caution and gentle fear kept her from having accidents.

Her friends are few, but are loyal and trusting. Friends that have stuck by her since childhood. As an only child, she spends her free time taking care of her aging parents. She feels closest to her cousin, Winnie. She wished she was closer to Winnie’s boyfriend, who was her ideal partner.

She was born in Taipei, at the Adventist Hospital. She’s only ever lived in Taipei her whole life. She couldn’t imagine ever living anywhere else.

When she’s angry, she finds a remote OK Mart convenience store and treats herself to packaged muffins. It calms her down, by forcing herself to chew slowly.

Her biggest fear is to be thought of as a fraud. She’s never told anyone before, but it’s a big, personal value of hers, especially when it came to the workplace. The industry knew her as an honest and trustworthy person. She’s never told her mother about her fear, because she knew her mother would laugh at her. (She considers her mother a fraud.)

Her secret is that she likes playing sock puppets alone at home. It’s a way to shake up her creativity muscles. Keep her socialized. Give her joy and giggles. She doesn’t really laugh out loud otherwise.

She has a boyfriend. She doesn’t love him. But he takes her out and treats her relatively well. There’s no reason to separate. In fact, they will probably get married someday. But if Winnie and her boyfriend broke up, Jocelyn would be there waiting. That’s who she truly loves. Even if he rejects her, it wouldn’t break her heart.

Jocelyn likes to always have snacks ready to go in her fridge. She hates warming things up in the microwave or oven, and preferred to slap on ham and cheese slices between two pieces of bread, with some mayonnaise and mustard, and munch on that instead of heating up leftovers. Her bedroom is plainly decorated but kept neat. Clothes put away, bedsheets straightened, and a small bedside lamp always on, to create the effect of a romantic, cozy room. She takes her trash out daily, even when the bins aren’t full.

She wears padded slippers in the house. She rarely has guests, but she still invested in a great deal of high-quality slippers for herself and potential guests. She wears a specific color depending on the day of the week.

Growing up, her parents’ home constantly smelled like cooking oil and soy sauce. They had friends over almost every day to play mahjong and gossip. Their helper was regularly in the kitchen cooking. For Jocelyn, in her own home now, she lights candles around the house, so she never has to encounter the smell of cooking food.

Before Lunar New Year arrives, she starts the spring cleaning process early. It’s one of her favorite times of the year. She gets to throw out old documents from work, which is very satisfying. Her childhood stuffed rabbit toy sits in her closet, collecting dust, and every year during spring cleaning, she gives it a wash, but she’d never throw it out.

It’s Saturday at noon. She sits at home with the lace curtains drawn, so there’s natural lighting but without the sun directly on her skin. She sips her hot coffee, black, in a porcelain off-white teacup, while journaling in her Moleskine notebook. She’s trying to lose weight, so she skips lunch on the weekends.

She thinks back to when she was nine years old, when she went to the zoo with all her cousins. There was a show for the seals, and they were taking volunteers from the crowd to play with them. Jocelyn was one of the special kids chosen, but she gave away her spot to Winnie, who was younger, and that’s how she justified it to herself, but this kind act came with a lot of pain and regret for Jocelyn. Even now, thinking back, she can only remember how bad she felt at that moment when she gave it away.

Tonight, she’s going out on a date with her boyfriend. She styles her office clothes in a way that’s appropriate for the weekend, because she likes the creativity of reusing her clothes in different ways.

For the next Writing Fiction task, the book gave a list of questions to consider when creating a fictional character. I followed the questions chronologically to write the above passages. I think it read nicely together, even though the questions were quite discrete.

1. What is your character’s name? Does the character have a nickname?

2. What is your character’s hair color? Eye color?

3. What kind of distinguishing facial features does your character have? Does your character have a birth mark? Where is it? What about scars? How did he get them?

4. Who are your character’s friends and family? Who does she surround herself with? Who are the people your character is closest to? Who does he wish he were closest to?

5. Where was your character born? Where has she lived since then? Where does she call home?

6. Where does your character go when he’s angry?

7. What is her biggest fear? Who has she told this to? Who would she never tell this to? Why?

8. Does she have a secret?

9. What makes your character laugh out loud?

10. When has your character been in love? Had a broken heart?

11. What is in your character’s fridge right now? On her bedroom floor? On her nightstand? In her garbage can?

12. Look at your character’s feet. Describe what you see there. Does he wear dress shoes, gym shoes, or none at all? Is he in socks that are ratty and full of holes? Or is he wearing a pair of blue-and-gold slippers knitted by his grandmother?

13. When your character thinks of her childhood kitchen, what smell does she associate with it? Sauerkraut? Oatmeal cookies? Paint? Why is that smell so resonate for her?

14. Your character is doing intense spring cleaning. What is easy for her to throw out? What is difficult for her to part with? Why?

15. It’s Saturday at noon. What is your character doing? Give details. If he’s eating breakfast, what exactly does he eat? If she’s stretching out in her backyard to sun, what kind of blanket or towel does she lie on?

16. What is one strong memory that has stuck with your character from childhood? Why is it so powerful and lasting?

17. Your character is getting ready for a night out. Where is she going? What does she wear? Who will she be with?

It was a new format and topic, where we only accepted ~40 people who had experience in marketing and had some familiarity with using AI to help with their workflows, then split them up into 5 groups with a moderator each to run group discussions. We had 2 rounds of discussions then opened up to a mixer afterwards for more free chats.

I was surprised that over 300 people registered for the event, so, unfortunately, we had to be quite selective and put most people on the waiting list. We read through the applications and let Claude help us weight the answers of why people wanted to come equally against their roles / companies.

It’s a good signal that this format and topic resonated with the community, so I hope to host more of these in the future.

We had interesting attendees join us. For example, we had an ecommerce founder who is doing 100% of their customer service with AI now. We also had a doctor who is trying to leverage AI more but getting pushback from his patients.

Overall, my observation was that there was split between how operators, managers, and owners used AI differently.

Operators are doing on-the-ground marketing on a daily basis, so a lot of the topics were around context hygiene across clients and keeping customers’ context separate without losing continuity. They also chatted about agent discipline, and one attendee recommended to start with fewer agents, then spin up more only when a task needs its own agent. We also chatted about how to leverage Claude Chat, Cowork, and Code in different ways.

I personally think I still need a human-in-the-loop when it comes to marketing. I don’t see AI being able to innovate on new designs or styles — it still needs a reference point.

For managers, I noticed that a lot of them were managing agents and team members, and the communication lines between them, rather than the work itself.

There still needs to be checkpoints in the automated workflows and processes from my personal experience. I still have to review everything before it gets published, because the accountability and responsibility still falls on me as the marketer and the one who ultimately runs these marketing processes.

For owners, it’s important to build an AI-native culture that employees actually embrace.

I’ve been interviewing more marketing people for my team, and I notice when people claim they are AI-native, they just mean they use AI to prompt on their chat interfaces, and haven’t actually made their workflows more efficient through AI automation or structures. We still have a ways to go.

Personally, my favorite recent new feature is the Slack integration. Andrej Karpathy joined Anthropic this year and championed this integration on X, where he got a lot of flak because people thought he was foolish from going from a cracked researcher and coder to supporting Slack bots. However, I think it’s much more than that. It’s putting the whole team’s context on the cloud. It’s actually incredible if implemented well, in my opinion.

For the format of the meetup, I thought the energy was great. Because the room was much more curated and intimate, there were a lot of lively discussions and connections made. People stayed behind late and went off to find a restaurant or bar downstairs to continue their conversations. I think this format is definitely repeatable for other topics.

I want to give a shout-out to my co-organizer Felice, currently working at Artificial Intelligence Foundation (人工智慧科技基金會) who also helped moderate, and the other 3 moderators Summit, 育軒, and Afore. Thank you also to Jessie from Appier and the AAMA team for the venue. Finally, a big thanks to the attendees who came and brought the good energy, as well as the 300+ people who applied that we couldn’t accept this time around.

Feel free to stay updated on future meetups by joining the WhatsApp channel, and can leave ideas or feedback in this Google form.

I don’t think it means we shouldn’t strive for it, but according to Darwin, diverse populations are more adaptable, passing on their “strong traits” to future generations, which isn’t really taken into account by our traditional definition of meritocracy.

I believe one of the reasons the US is considered so powerful and successful today is because it has been, and continues to be, the most diverse place in the world. And because of that, meritocratic ideals should evolve to capture how diversity is a variable that needs to be baked into what that really means.

party in the USA

There is no such thing as a perfect system. Many people have fought for the US to be more welcome to diverse people, and while it’s better in many ways today, it continues to be something worth fighting for. This “fight” isn’t just between white people and black people, or locals vs immigrants, or rich vs poor, etc. It’s a fight against a system that, whether intentionally or not, gets built to sort out one out-group from the rest, because it means at any point, the in-group could become the out-group, and the me against them mindset will work to tear people and societies apart.

During the Black Lives Matter movement, I wrote about my university friendships back in Cleveland on Instagram. These women remain, to this day, my closest friends and my strongest support system. I think I was a little shit back in high school, but I must have done something right to be randomly placed as roommates with Amanda, the most amazing woman who changed me to become a much better person. I regularly think back to my university days hanging out with Amanda, Chino, Ify, and Peace, and how I was at the top of my form in every way (physically, mentally, emotionally, and spiritually) because of the environment we created for each other.

Since we were such close friends, we were always together on campus, which allowed me to witness some very strange things that happened to Black people. For example, Black students notoriously avoided Little Italy, a neighborhood where Italian-Americans populated, because of how they were treated. Once, when a group of friends went to try an empty restaurant there, the server brought out a water jug that had a hole in the bottom of it, leaking water all the way as she walked over and then all over the table when she smacked it down. They wouldn’t take any food orders from my friends, so they left. I never went to Little Italy due to the hostility.

When I moved to Las Vegas to work at Zappos, I was seeing my ex-boyfriend at the time, someone I had dated since university and had moved with me from Cleveland. For some context, my ex is Black. We once went to a downtown Las Vegas casino restaurant, and even though the main section of the restaurant was relatively empty, save the few white people eating and drinking there, they took us to the back section, a separate dining room, that was not decorated (the walls were bare), right next to the kitchen doors and bathrooms, and it was completely full of people, but only non-white people. I was shocked to experience this treatment in 2015, especially in such a tourist town like Vegas. I was uncomfortable that people passively accepted this situation. So when the Black Lives Matter movement happened, I personally understood how necessary it was to speak up.

While there are laws that should protect against such aggressive and discriminatory behaviors, the pattern still exists, and I’m sure everyone who’s lived in the US has experienced some form of it. It’s one of the reasons I moved back to Taiwan, because as a non-American myself, I knew I was fighting an uphill battle that I would not be able to win, given that I am considered second-class in every way (non-American, a woman, introverted, with schizophrenia.) I feel privileged to be Taiwanese and to be able to build a healthy life here.

riding the momentum

I bring this up because a couple of years ago, I read this biography of Martin Luther King (MLK) Jr. by Jonathan Eig that gave me a lot of insight into MLK Jr.’s peaceful revolution, and one of the few that actually made a difference for the people and communities they were fighting for. Its success is worth studying for communities anywhere.

For example, I thought the famous Rosa Parks story was made famous accidentally. For those that aren’t aware, back in 1955, the US was still segregating public spaces between white people and everyone else. Rosa Parks, a Black woman, got on a bus and decided to sit outside of the colored people section, and refused to move when they demanded her to. While her refusal wasn’t staged, the National Association for the Advancement of Colored People (NAACP), formed in 1909, made a conscious choice to build the boycott around her case in a strategic way, highlighting her as someone that people would sympathize with.

violence against the non-violent

I had also read the Malcolm X biography, who had a very different path compared to MLK Jr., but the interesting thing is that while Malcolm X started off as pushing more violence against aggressors, he eventually chose the more peaceful route, and then he was assassinated. MLK Jr. himself was quite a peaceful protestor, and he was also assassinated. Although their methods in the end pushed for peace, the outcome was still violence.

Reading these stories gave me some conspiracy theories to think about as well. For example, the Director of the FBI at the time, J. Edgar Hoover, was known to loathe MLK Jr. and believed he was the biggest threat to the US at the time. He set up surveillance, wire taps, and a political campaign against King. There is solid proof today that the FBI ran a sustained, aggressive campaign against King.

In 1963, President John F. Kennedy (JFK) was assassinated, and his VP, Lyndon B. Johnson (LBJ), became President, at the same time as King was moving his peaceful revolution along. Kennedy had a working relationship with MLK Jr., and in my personal tinfoil hat theory, I think Hoover had Kennedy assassinated because LBJ was easier to control. I think Hoover disliked JFK, and probably thought (correctly) that he could get to MLK Jr. more easily through LBJ. Fun fact: LBJ was married to Lady Bird Johnson, who happened to become very close friends with Hoover before the JKF assassination. Given that she became the First Lady of America after JFK’s assassination, I think it’s all connected.

MLK Jr.’s family all died soon after, one by one; first his brother who drowned, then his mother who was shot. In my opinion, they were all assassinated. In 2025, a portion of the files have been released, but the full surveillance tapes are sealed until 2027, so the material is still hidden.

In the end, MLK Jr. was more than a martyr. He showed the importance that peace is something worth fighting for, not something to sit around and wait to appear. Diverse societies that sort people by group are unjust and structurally unsustainable, so his fight is everyone’s fight. What Black Americans are fighting for here is for all of us out there who believe in truth, justice, equality, and freedom.

From my personal observations, web3 people happen to be some of the fastest adopters and strong users of AI. People joked that web3 folks jumped from one trend (crypto) to another (AI), but I think it’s more nuanced than that. Obviously, I am also a web3 person, so there might be some slight bias, but I do actually believe that people who were willing to enter the web3 space as early tech adopters are also the ones most likely to leverage AI, another new technological advancement for human society.

This time around, because of the nature of web3 people as well, I wanted to keep it more casual in terms of the vibes, and less like a lecture. Even the speakers sat quite casually at a table to field questions during the Q&A afterwards. More importantly, I tried to find a venue that could create that vibe, going for one that made it feel like we were at a large cafe, so attendees could co-work while listening to the talks and meet new people more naturally.

Although it was raining, the turnout was great. A lot of people stayed behind to chat with each other, to the speakers, and to me. I even found out that one of the attendees, who is a university student, had been following my posts, where I sometimes share about the struggles of trying to achieve greatness, whether it’s my personal stories or recaps of stories I heard from other founders, and she started vibe-coding because she realized it is a journey, and not something some people are just born with. It was very validating to hear that.

The next meetup I’m planning is going to be less about demos and more about splitting everyone into smaller groups with moderators to discuss how marketing folks use Claude to power their workflows. We are keeping it intimate at 30-50 people. I noticed there is a great deal of demand for this one after launching it, which makes me wonder if it’s the topic (marketing) or the format (small group discussions) that people are really interested in. Probably a mixture of both, so I will try another marketing-related one at some point, and also this small group discussions one again.

Feel free to shoot me some ideas if you have any or want to collaborate~

In the industry, it’s used more like a noun: we need more compute.

“Compute” is shorthand for raw computing power, which is the capacity to do calculations, and it’s a quantity you can buy, rent, stockpile, or run out of.

It’s similar to how we would talk about water or money, so in a way, it’s like a currency or resource, and this is indeed representative of how it’s used in the industry. People want more compute.

So when someone says a model “took a lot of compute,” they mean it ate a huge amount of processing capacity.

And in our AI world, roughly, the more compute we put in, the more intelligence we get out.

When I look online, I see a lot of great, complex, technical content about this industry, but it’s hard to follow as a beginner.

I decided to start documenting my learnings here from 0, alongside my regular writing, and for those who are starting from scratch in this space, I welcome you to join me on this public learning journey.

While I’m doing this partly because I want to get better in my role, I’m also doing this partly because I’m Taiwanese and I want to understand the industry that’s propping up our GDP, and partly because it’s a big topic now, so why not try to learn about it in a self-improvement kind of way.

I’ll start simple and hopefully keep getting more advanced (or as advanced as I can be as a non-technical person). I’ll keep each post quite short and easily digestible, as that’s how I’ll be learning about it too. I’ll post regularly, but each post will take less than 5 mins to read.

Feel free to subscribe to my new newsletter about this topic » head in the cloud.

When I was at AppWorks, one of my favorite initiatives was the internship program, and every batch without fail, I would make a big effort to interview and bring on an intern to mentor. The Zappos internship changed my life and I wanted young people in Taiwan to be able to experience something similar for their own career trajectories, and I met so many great people that eventually became some of my really close friends, including Hazel.

When we interviewed Hazel, her English was quite poor, and she didn’t have a particularly standout resume. My coworkers at the time were uninterested in taking her, and I had a choice between her and another candidate from NTU doing Finance, someone who seemed much more impressive at first glance. However, I picked Hazel in the end, because she had such an optimistic and brave attitude, and I think it’s one of the best decisions I’ve ever made in my career.

Hazel did not come from a privileged background. She worked really hard at AppWorks, and after graduating, went through a lot of career ups and downs, moving from an investment intern to an HR function to being a self-taught data analyst at MoBagel, JKO Pay, and now at Coupang. She’s also tried to build a data community on the side, but it was hard to balance with demanding workloads, so she discontinued it after a while. But she still keeps in contact with many of the initial members and her mentees. Seeing Hazel’s growth makes me really proud and bullish about young people in Taiwan, and she’s also one of my role models.

In recent years, Hazel got connected to Professor Amy from NTHU who teaches technical business courses like data analytics and technological marketing, who invited Hazel to do guest lectures for her data courses. This year, Hazel recommended me to speak at Professor Amy’s class, which was an honor, and not something I’ve done before. I mostly talked about using data to make business and marketing insights, and I think if I were to do it again, I would use a Slido throughout to collect questions, as some of the students may have been shy to ask questions out loud in front of the class. However, Professor Amy did notice that many students that usually don’t speak up ended up asking questions, which was a good signal.

I won’t rehash the marketing concepts here, since I’ve talked about some of them before when I shared about my marketing freelancer collaborations. A few weeks before the lecture, I had put together a Google form to collect answers from the students that I used during the presentation, and I wanted to share some the things they thought about when it came to marketing and the future. There were no right or wrong answers, but it was interesting to see what they thought.

what does marketing mean to you?

I asked: what does marketing mean to you? The most answered was “awareness,” followed by “understanding people and delivering value,” to a close third of “convincing and persuading people to buy things.”

These are all related to marketing, but the one I personally prefer when I think about doing marketing is the second one: understanding people and delivering value. To me, awareness is just one part of the user journey, right at the beginning of a campaign, and convincing people to buy things is in the middle of the journey when they convert to becoming customers. In the end, if you understand your users, you can deliver true value to them, which will increase both their awareness and conviction to buy.

data in marketing

I asked: what matters most for marketing success vs what matters most when you are making a purchase? For many students, creativity in marketing was rated as the most important, with data far behind, and experience last. However, when they make purchasing choices, doing research and having data becomes the most important. This mismatch is common when thinking about marketing early on, when you have less experience. I thought the same as well. For myself, I would rank experience first and then data.

what brands are effective at marketing?

I asked: what are your favorite brands, what commercial stands out to you, and what brands or products have you recommended others and why? When I coworked with WeMo Scooter last year, the marketing team asked NTU’s data class to survey their student population to see what resonated with them, and it seems consistent across universities in Taiwan that Apple is the top answer. There’s a certain aspirational vibe and cool factor in owning Apple products.

In terms of ads that are memorable, even though some of them were funny or gross or feel-good, it didn’t necessarily translate into people becoming customers. They just happened to be memorable. However, I liked the answer a student gave about Pocari Sweat, where the ad reminds you of how you feel when you drink Pocari, and whenever you want to feel that way — refreshing and summery — you’ll be more inclined to drink Pocari.

In general, if you see the brands or products that students actually recommended to others, it focuses on the effectiveness and cost, the latter being highly relevant to students who have lower purchasing power. It had nothing to do with display or visual ad campaigns.

the future of work

I asked: how worried are you about AI affecting your career, how many of you use AI regularly, and what does your current online presence look like? Most students were pretty worried about AI when it came to their future careers. To be honest, this might be common across the board even for working adults. It’s especially telling that all of them are using AI, and the majority of them are using it regularly.

There’s a lot of advice out there that’s already being said, like having taste and judgement, which isn’t something AI has the ability to do currently. For me, real relationship-building is something that will be extremely hard for AI to truly replicate, and one of the ways is to develop an online presence and connect / communicate with people digitally. It’s been a big reason why I started writing more, and learning that it’s a 2-way communication channel. I learned this originally from my previous boss Jamie, who started his Taiwan career journey by publishing blogs online. It’s also something that’s easy for students to start right now, and I advised them to focus on informational or educational content first if they don’t know what to write about.

I planned an interactive activity, using Claude to generate a fake Momo data report.1 I asked the students to look through the data and tell me what they would put on the homepage of Momo’s website tomorrow and why.

Again, there were no right or wrong answers. In marketing, you’d have to test these decisions before you know if it was the best thing for the business. I used to do this everyday at Zappos, except it would be for deciding which products to give more budget to when running Google Ads.

Feel free to download a copy of the data report and play around with it! And let me know what you’d put on the homepage, and I’m happy to explain how I would do it myself » Google Sheets

Finally, I would like to make a plug for Amy’s classes. She’s a great person and really invested in her students’ growth. If there are any founders or professionals out there who would be willing to share about how they think about data, marketing, anything business-related or career-related, please let me know because she’s always looking for great guest speakers or ways to collaborate so students can get real-world experience.

I first encountered Replit after I left my last full-time job and had more free time to do some self-learning. I decided to try an online coding course (which I never finished after having a baby and starting to work again hah) but it helped me feel more “productive” and the course I took had us use Replit.

Replit, as the company, was founded in 2016, though the open source side project dates back to 2011. Their revenue was essentially flat until September 2024, when AI models could enable agents to work well, and it jumped from US$ 2M ARR to $70M ARR within 6 months. The decade-long, continuously-building persistence is admirable.

The Social Radars podcast interviewed the co-founders Amjad and Haya, who also happen to be husband and wife, similar to Paul Graham and Jessica Livingston. I find this to be really inspiring given that my husband and I started off as coworkers.

Here’s some of their story in building Replit:

Amjad’s background is in computer science, but he felt that learning programming was starting to get really difficult in the 2000’s where you had to study it in school in order to code. He mentioned that it wasn’t always that way, for example, when Lisp was the common programming language, it was easy to pick up and make a website, but over time, it got more and more complicated. He modeled Replit off of how that used to be; when you could go to a website with a console and start programming straight away.

Back in 2008, he found it to be quite surprising that nothing like this existed yet. It took him a few years to make the application more useful and open sourced it to Hacker News in 2011.3 He was dating Haya at the time and recruited her to help him build it, since she was studying to be a designer.

Amjad originally had dreams to be a founder, but then he got a really good opportunity to work on CodeAcademy, and when he got the job offer, he decided to propose to Haya. Haya mentioned that getting engaged was a very practical decision where they both knew that they were unlikely to break up, so it made sense to get married.

In late 2013, Amjad wanted to do something more technical, as he was mostly more on the leadership and hiring side at that point in time, and he had also seen Facebook do some very inspiring work. Zuck was talking about how everyone was going to be connected and the internet was going to change the world. Amjad himself had seen how the internet had changed his life and democratized the world, so he joined Facebook and got back into dev tools, and he realized that the tools were still so bad even for a big company. In 2016, he started getting the itch to do something new.

At the same time, Haya was trying to chase the American dream while they were both living in NYC, but she was also competing with the best designers, since it was NYC. She kept practicing her trade, finding different gigs, doing a stint in the MENA region, joining bootcamps, working on side projects, but she felt she was barely scratching the surface. Amjad encouraged her to stop searching for a job for a year and they both decided to try to figure out what it takes to become a designer at Facebook. All the designers at Facebook had side projects that were very interesting, so they focused on what Haya could work on. A fitness app? A lifestyle app? But wait…what happened to Replit?

Haya had really enjoyed working on Replit before, so she decided to continue the project, talking to users and asking them to do interviews and surveys, and she found out there was so much love from their small base of users. While she was working on Replit, she was interviewing at different companies, and one of them asked her why she wanted the job if she was so into Replit, and she had an epiphany moment to commit to it full-time.

Haya and Amjad both decided to make the plunge into working on Replit, so they came up with a master plan to go all in. They initially applied to YC as an education startup but got rejected. They raised elsewhere at US$500K, which helped them reach profitability, and then in late 2017, they got a DM from Sam Altman telling them that YC was interested in them. Amjad laughs about it, asking Sam why they they weren’t originally accepted, but anyway, Paul Graham had found them on Hacker News in 2017 and asked Sam to invite them into YC.

Amjad and Haya had always wanted to evolve the product outside of just education. Amjad wanted to build a great programming platform, and while it should teach people, he wanted people to be able to build real things with Replit. Joining YC gave them the license to go big, because of the energy it gives. The orange aesthetics of the coworking space, the way people talk making founders feel like they’re doing something meaningful, constantly being surrounded by great people. It also felt competitive, which kicked things up a couple of notches for Amjad and Haya.

A big part of developing Replit was figuring out how to get it in the hands of non-technical people. How do we get non-engineers to understand it and for it to be simple enough to start building?

Another interesting variable that made Replit strong was its security. It isn’t something most people think about when it comes to using their product, but to them, it was so important that people could build without the fear of fraud, hacks, or malicious behaviors. They referenced Steve Jobs’ “connecting the dots backwards” concept when they think about this. Amjad has a security-related background, and even though there are other competitors now in the same space as Replit,raising big rounds with a large team, what makes Replit stand out is that they built in security features from the beginning. That’s also something they liked about their YC experience, where those on the outside know what’s hot and what to copy, but they don’t actually know the secret sauce to its success, for example for Replit, it was their security.

Amjad and Haya also talked about working as spouses. Of course, there are many challenges, like in the early days when they would ping pong responsibilities and couldn’t really rest, and in general, investors have a bias against couples as founders. That’s a reason they also really liked YC, because Paul and Jessica understood. Being married and also being co-founders tells you a lot about how decisions are made, whether at home or in the company. It’s like a second kind of marriage: will you leave if things get really rough? Sometimes, Amjad and Haya even had to bring their kid to work, because they had their first kid during the pandemic while they were scaling like crazy and didn’t have access to childcare. It was the toughest period of their lives, because they were growing fast but were still broke.

They shared about how some of their users are now using Replit, those who wanted to be entrepreneurs but didn’t have the technical background to do it, and now with Replit, they can. They have a user who built a product to US$1M ARR even though his background is in photography. Another user was a PM at Microsoft and built a PowerPoint alternative with Replit, then quit to do that full-time.

I thought their story was cute. It had a lot of elements I learned about in the startup journey, like not giving up, like healthy co-founder relationships, like finding a small group of users that love your product (rather than a large base who’s not so committed).

I never finished that coding course I started. I do plan to someday, even just to understand the theoretical concepts of working with computers, but as a lifelong procrastinator, this will take some time.

For example, one of her pieces A Cloud A Day allows site visitors to share their location and she had connected it to a weather website API which would generate a cloud based on the date, location, and current weather of that location onto the page.

I do some art occasionally, like ink painting, but I wouldn’t really consider myself much of an artist. However, I have a lot of artist friends, who went to school for art, have a deep passion for creating art, and I’ve always been fascinated by their minds and how they look at themselves and the environments around them.

Especially now with AI, we hear a lot of vocal backlash from the art world online about how AI is terrible for a variety of reasons, like copyright infringement or how it fares against human-only skills, and so on. However, there’s definitely a good portion of artists who are embracing this tool and leveraging it to create more art. I also think that AI helps democratize a lot of abilities that were previously unavailable to a lot of people, something that has written about before.

I chatted about this idea with Nicci and asked her to be one of the demo speakers for a Claude meetup where we can focus the topic on how artists use Claude and AI to create new forms of art. I also pulled in another friend from middle school, Jade, who has been in the art industry her entire career and currently at the Zao Wou-Ki Foundation, who would be able to introduce me to some local artists. That’s how I met 2 of the other speakers, Sam and Rong.

For Sam and Rong, who previously didn’t have much experience with AI, I gifted them Claude subscriptions so they could try creating a few works with AI. Sam ended up sharing about his process of turning his still paintings and prints into animations, and Rong talked about how she learned mathematical concepts through emotions to create interactive patterns.

Because of my personal interest in art, I knew a few friends who owned or supported art galleries around Taipei, so I reached out to Shang who owns a beautiful space in Neihu called Koshi Gallery. I thought it would be quite suitable and cute to host this artist-related meetup at a gallery.

A few weeks into the planning, I remembered a new media artist that I had followed during the NFT days, Che Yu Wu. I decided to cold-email him to invite him to attend our meetup, and to my surprise, he replied and said he was interested in coming. So I pushed a little more and asked if he’d be willing to demo, and he said yes as well. How exciting! Che Yu’s talk focused a lot on creating life forms with AI and “growing” them, like pets. He applied this thinking to his most recent viral project, Taiwan.md.

Due to the limited space at the gallery, we had to be selective about who to invite. In the end, I optimized more towards people who had backgrounds in art and design, and it was great to meet so many different kinds of people who were Claude users and using Claude to make art. I would definitely like to try this format and grouping again. I think Taiwan has a pretty cool underground art scene in this specific intersection with tech.

Thank you so much to my planning crew Nicci and Jade, to our great demo speakers including Sam, Rong, and Che Yu, to Shang for the gallery space, to our attendees, many of whom stayed after to engage more with the speakers and each other, and to the Claude Ambassadors community for supporting this meetup.

If you’re interested in reading about the demos of each speaker, you can check out this Google doc.

Forking is a known concept in the blockchain industry. Ethereum was probably the first to popularize it, when the DAO was hacked back in 2016 and US$ 60M was drained. It was early on in Ethereum’s journey, so a hack like that was existential to a young ecosystem. The community ultimately decided to fork, while a minority stayed on the original chain (ETC), which continues to exist today.

ETH became the dominant chain by every meaningful measure: developer activity, TVL, cultural relevance, and adoption from non-crypto-native industries. It validates my assumption that forking is not as bad as it might seem. I think it shows a willingness to course-correct, even at the cost of “purity,” which is what saved the Ethereum ecosystem in the end.

the case for forking

I started thinking about forking when I was part of several DAOs a few years ago. In one of the DAOs I was in, there were some differences in opinions about what direction the DAO should go, especially between the founders vs the people running the operations. I was gently trying to suggest a fork, but I didn’t think it was well-received, so I didn’t push for it much.

Everything has been resolved now, but I really feel like things could have turned out better if we had tried a fork. I could see the social dynamics were difficult to overcome when the topic came up, because there was a fear it would fragment the community, a perceived betrayal against the founding vision, the coordination cost, and an admission that things weren’t working out as planned.

Working in the VC industry before and seeing what so many startups and founders went through, I felt differently about it. It’s common when things don’t work out in the startup world, and sometimes that leads to shutting down the company, pivoting the business, or the team separating and doing their own thing. A recent historical example is how Dario Amodei left OpenAI to start Anthropic, resulting in better options for consumers and the world.

Instead of seeing forks as a failure, I see them as a mechanism for resolving disagreements when consensus can’t be made. I think it’s better than the alternatives: forced unity with disgruntled members, slow decay, and resentment. In the end, I should have approached it less strategically and more emotionally.

A fork also lets both visions be tested, and reality picks the winner. I feel that’s the most fair and in-line with how the free market works. It also shows a natural progression of how things play out.

To be fair though, sometimes forks don’t turn out so well. In blockchain, Bitcoin Cash forked from Bitcoin in 2017 over the block size debate, and then forked again into BSV. It became more fragmented instead of thriving, and lost cultural relevance (in comparison to Bitcoin). Forking isn’t always the best approach nor does it always work out.

If we apply forking as a general concept beyond software, I think Taiwan as we know it today was a fork from the PRC trajectory after 1949 politically, and a fork from the shared Chinese heritage that developed differently here culturally.

We are ~100 years in, and while it’s not perfect in many ways, I would consider us a thriving country and peoples. Politically, we evolved from martial law into a full democracy with peaceful power transfers and high civic participation. Economically, we transformed from an aid-dependent island into a tech superpower, home to TSMC and a semiconductor industry that now underpins the world’s digital infrastructure. Culturally, we’ve nurtured a distinct, open identity with strong free speech, an active civil society, and a livable, innovative society that continues to trend upward.

Evolutionary biology shows how life systems adapt by branching out over time. Forking can be painful, but I think that highlights the seriousness of the decision, and could be seen as a healthy progression rather than a tragic one.

Eigenlayer’s founder Sreeram Kannan mentioned that forking is a superpower. One thing I love about the spirit of being in the blockchain ecosystem is that most of the builders are overturning traditional mindsets of what is “right” to being more open-minded about what’s possible. Even now, meeting new crypto people who are still building, makes me stay bullish that it’s about the long game.

I’m writing this for the primary purpose of highlighting the speakers. I would also like to share some of my takeaways and learnings from the panel, and I hope it’s interesting to those who want to hear more about what Taiwanese founders are building.

Thank you to Guillaume and the Headline team for the opportunity. There was a great mix of attendees at the meetup.

If you’re in the Taiwan startup circle, you’ve probably heard about MetAI already. They build digital twins and simulations for factories and the manufacturing industry, focusing deeply on the needs of physical AI. They are also the first Taiwanese startup to be invested in by NVIDIA.

Daniel is the CEO of MetAI and has an interesting, non-linear background that led him to where he is today. He’s had experience in engineering, in finance and investments, in blockchain, and in VR, but deep down he’s really a product guy, which is why leading MetAI is exactly the right place at the right time for him. After experiencing how a factory runs, he realized that one small mistake could become very costly because of the equipment needed, and that’s what MetAI has set out to help solve.

When he talked about how he works with domain experts as his clients, he admitted that being humble is a strength, and he learned this the hard way. When he first started talking to potential customers, he mentioned he was arrogant and assumed people would buy into their tech since it was so good. He soon realized how that was the wrong way to approach talking to customers, and even though the technology is truly impressive to him, he spends more time getting to know his customers and building a warm relationship before any assumptions or expectations are made.

I thought Daniel’s sharing was rooted in a lot of conviction in his product, his team, and himself. Looking forward to seeing more of what MetAI will do these coming years.

Agnes has been building GliaCloud for around a decade now. GliaCloud is platform that helps advertisers and businesses tell stories through video creation. Agnes is the co-founding COO, with a background in finance, and she mentioned that when they got started, it was mostly just her and a bunch of engineers. At the time of launch around 2015, no one believed that AI could do anything, let alone create videos. Luckily for them, the managing partner at Headline, Akio, could see their vision and invested in them early. Now, GliaCloud serve thousands of clients with their video creation needs.

Agnes gave very practical advice, talking about how their customers are ultimately still focused on how GliaCloud can drive business results. Even though AI can now make endless amounts of content and in different formats, it doesn’t mean those kinds of videos will help businesses get more customers in the end. That’s where GliaCloud comes in, so there’s still human elements that can’t be replaced by purely AI.

There’s a funny meme that designers were sharing, about how clients were bragging that with AI they can do everything without designers now, without realizing that they still don’t know how to prompt correctly to get the design they want, just like they don’t give proper context to designers about what they want the design to look like. Agnes is no stranger to this kind of mindset and sees both sides, from the design perspective as well as the business perspective, which is why she’s so suitable to help bridge the gap of what businesses really need.

It was great to have Agnes on the panel, where you could tell she thought deeply about each of the questions before answering. She had a grounding presence and shared very realistic and specific takeaways for the audience.

I met Bill when he went through AppWorks’ accelerator program as PreTeeth AI, which he pivoted to Dentscape afterwards. Dentscape is a physical AI solution for dental clinics and labs to create new teeth for patients who need them.

Initially, he focused on serving the end users directly, supporting them when they felt they needed new teeth for practical or aesthetic reasons. He talked about how, like many founders, they got to a point where they were running out of money and the demand wasn’t enough to sustain them, even though their tech was good and it was actually solving a real pain point. He ended up connecting with the largest dental clinic in Taiwan who told them that their solution would really help with their needs, and even invested in Dentscape once they focused on delivering value to dental clinics and labs instead.

Bill also found out how dental technicians have decreased by 50% over the past decade because of the grueling work hours, the average pay, and the difficulty of the job. This is despite how the dental industry continues to grow steadily as people have more awareness and more spending power to do work on their teeth. Therefore, Dentscape offers a much needed solution to the industry to keep it going, starting in the US market. Their ask at the end of the panel focused on finding more great sales reps with international experience to help grow the team.

Each of these founders shared an important takeaway: solve for a real need. This leads to driving real business results for their customers, which is how to really deliver value. I learned some new things and insights into specific industries chatting with these founders. I hope the audience did too.

IVS is the largest startup conference in Japan that’s held every summer by the Headline team. This year, it’s on July 1-3 in Kyoto and they’re anticipating over 20,000 people to go through it. If you’re a founder who wants to land in Japan, this is a great event to get involved in » https://www.ivs.events/en/

One of the reasons I still want to write about it, even though it wasn’t great, is because I know I am a big consumer, whether it’s through books or film or video games or food or whatever else. I previously wrote about how I saw consumption as Yin energy, and I was making more of a concerted effort to balance that out with Yang energy through creating more aka writing. So I still want to write about these games, just not dedicate an entire piece to each of these games.

Here is a shortlist of mid games I played in recent years:

Biomutant: This game seemed more like a game for kids, but I thought if younger gamers were the target audience, it wasn’t too bad to experience what an open-world hack-and-slash game was like. There was no real story though, so I found it quite boring.

Monark: It had good reviews and seemed quite interesting with the story and the Persona-type vibes, but I felt there was a huge learning curve for me to actually enjoy playing it, so I didn’t end up finishing it. I play games at night after I don't have much brainpower to work anymore and my son has gone to bed, so it's more to let loose, and I don’t have the energy anymore to learn new games that require a ton of strategic thinking.

Gotham Knights: it was a beautifully-made game but it was pretty boring and repetitive with no real storyline. The premise wasn’t bad but it was a bit corny (Batman wasn’t around and his team had to live up to his legacy to save Gotham). The fighting was realistically violent though.

Launching a game seems pretty hard. Once it’s out there and the experience is poor, it seems it can be hard to make the experience or reputation better. I know that some gaming studios will let gamers have early access to collect feedback before the big launch, and it really made me curious if Guardians of the Galaxy did that because I can’t believe they launched it with automated running. You could tell from the very beginning how annoying it was.

These mid experiences made me reassess how to pick games to play, especially given that I can’t play very often anymore. Initially I just chose more loosely, because I have this irrational fear of finishing a great game and not being able to find a better one afterwards. I should just YOLO it tbh.

For this meetup, I reached out to the Web3 For All community, specifically my former coworker Liying Wang, who is a lawyer by trade and founded this community because of her passion around technologies that change people’s lives. Although we capped in-person attendance at 50 people, so that audience members could have deeper discussions during the mixer portion, we shared a Google Meet link since we still had decent demand from people who wanted to learn. Thank you to all the attendees, both in-person and online, for joining and engaging with the meetup.

There was a great mix of in-person attendees, ranging from PhD policy researchers to someone working in the UN on AI governance that just so happened to be passing through Taiwan at the same time as the meetup. At the beginning of the session, we went around the room and each attendee said one word to represent themselves or what they were thinking about. Trust. Security. Allegiance. Courage. Resilience. Legal. Representation. Open source. Builder. Community. It was insightful to hear such diverse contexts and their relevance to the topic we had planned.

In terms of the main sharings, Peter Cui set the tone with framing the ongoing case between Anthropic and the US Department of War, Nicole Chan brought a governance and policy perspective, and Anshuman Prasad gave a practitioner’s view from inside the defense tech industry. It was informative to hear takes from different angles, and as an individual citizen who doesn’t have to deal with these questions on a daily basis, it gave me a lot to think about when it comes to what is “right” and what I would think about in their positions. I’m very grateful to our wonderful speakers for sharing their perspectives.

I think I would like to do more of these kinds of philosophical discussions in the future when it comes to Claude meetups. AI ethics and governance is something that we should all be aware of, given than most of us are starting to use AI regularly, and especially how it affects us as citizens of Taiwan. I’m happy that the Claude Ambassadors program is supportive of us trying these kinds of meetups as well, so a big thank you to the Anthropic team for giving us this opportunity to do something new.

Overall, I had a great morning learning and thinking about important things that I don’t think about on a daily basis. Thank you to my co-organizers Peter, Liying, Jocelyn, and Mashbean from the Web3 For All community and thank you to g0v for the venue and vibe.

Here is a brief write-up of what each speaker talked about, and some personal thoughts about the topic.

Peter: the ongoing lawsuit

Peter walked us through the timeline of what happened between Anthropic and the US government. From 2023 to 2025, Anthropic was a Pentagon partner. Then, starting in fall 2025, a dispute emerged over the terms of deployment, specifically that Anthropic refused to include an “all lawful purposes” clause that would have given the Department of Defense unrestricted use of Claude. By February 2026, the government designated Anthropic a supply chain risk and pulled its contracts, which prompted Anthropic to file suit on March 9.

A quick breakdown of opinions between the two parties:

• On Anthropic’s side: First Amendment retaliation for protected speech, Fifth Amendment due process violations, and APA / ultra vires claims that the administration exceeded its statutory authority.

• The government’s position: national security deference, broad procurement discretion, and the argument that this was about safety and procurement, rather than about punishing free speech.

From my limited understanding of US laws, some of this seems like a language conflict. I wonder if Anthropic just used stronger guardrails against things they wanted to fight against, like mass surveillance, that “all lawful purposes” would still be allowed. It feels like they are fighting two separate fights because of these technicalities. Of course, I don’t know the true intent of either sides.

Peter also walked us through the coalition perspective, as there were free speech and constitutional advocates including FIRE, EFF, and the Cato Institute, the tech industry, former military and national security officials, Catholic ethicists, and a law professor from the University of Minnesota who were supporting Anthropic. Some other groups more focused on human rights like Access Now, CCR, and others, didn’t support either and criticized both. The case is still live and the litigation is running across three courts simultaneously.

Peter also explained the OpenAI vs. Anthropic governance comparison about who can actually say no when the pressure comes. OpenAI’s structure allows leadership to accept residual risk and their commercial scale drives deployment. Anthropic’s LTBT board structure and maintained red lines meant they were the company that said no. But does it matter saying no if you just get replaced?

I don’t think there’s an easy answer. Personally, I think it still matters to be able to say no when you believe strongly about something, even if you’ll just get replaced. It seems like a lot of other people agree, because in the week that Anthropic announced they weren’t working with the Pentagon anymore, Claude sign-ups surged to almost 200% growth since the start of the year, and “QuitGPT” started trending, leading to almost 300% uninstalls experienced on just one day.

It’s hard to decide if Anthropic did the right thing or not. If you own the best technology, wouldn’t you want it to be the default in governance over difficult decisions? You wouldn’t want that to go to a less moral person, right? With great power comes great responsibility. And yet, deciding to hold their ground about how their AI can be used was the act of power and responsibility that Anthropic believed was right. And I would hope that if I’m ever put in a position like that, I could do what I believed to be the right thing, even with a lot of external pressures.

Nicole: from responsible AI to trustworthy AI

Nicole is the former chair of Taiwan’s National Communications Commission, and currently the chair of the AI Technology Foundation and the Taiwan Digital Trust Association. Her perspective is from a governance framework.

She believes AI has moved from an experimental technology to core corporate infrastructure, while governance is still catching up. The transition she cares about is going from AI that’s responsible to AI that’s trustworthy. Responsible AI is about the ethics of the actors building and deploying systems. Trustworthy AI is about whether the systems themselves can be institutionally verified and socially trusted.

She laid out how different governments are approaching this.

• The EU went first with the AI Act: risk-tiered, mandatory, with enforcement mechanisms and the European AI Office to coordinate.

• Korea followed with its own AI Basic Act, set to take effect in 2026, positioning itself as the second country globally to pass comprehensive AI legislation.

• Japan took a softer approach: the AI Promotion Act emphasizes facilitation over restriction, relies on industry self-regulation, and appointed Chief AI Officers across government ministries.

• Singapore operates without a unified law, using sector-specific guidance from bodies like IMDA and MAS.

• The U.S. under the current administration has framed its AI policy around competitive advantage and national security rather than risk classification or social protection.

Nicole has been thinking about these two questions lately: when AI becomes a survival tool, will humans still pay the ethical cost? And when AI becomes war infrastructure, who has the right to draw the line?

She closed with what she called digital constitutional transformation. The confrontation between Anthropic and the Pentagon is not just a contract dispute or a First Amendment case, but also a function for decisions that democracies haven’t made yet. It’s an existential question for nations, and defining what AI governance looks like will be a core part of this global paradigm shift.

In a democracy, it’s noble to try to get everyone involved in deciding what to with something so encompassing and impactful. From a world power perspective, I can see how governments feel about moving fast vs slow when you have to wait for everyone’s opinions, and we do elect our representatives to make these kinds of decisions on behalf of us, but I’m grateful we are still able to voice our thoughts, and even have an impact in decision-making if we feel it’s extremely harmful. That is a part of capitalism that I really believe in, when consumers can “vote” with their wallets, like people switching from ChatGPT to Claude.

Anshuman: the framework > the fight

Anshuman’s background spans Palantir, General Atomics, Protocol Labs, and his current role at Valinor Enterprises, a defense tech company. As someone who has actually built AI systems for military use, his take was more about the operational reality rather than governance philosophy.

He set the tone of his talk with some history: the military has always adopted the best available technology. Expert systems in the 1950s and 80s. Precision targeting and ISR fusion in the 1990s and 2000s, which became the foundation of Project Maven. ML-driven drone operations in the 2010s. Now: foundation models, edge inference, and autonomous systems at scale.

Anshuman gave us some context about what defense AI cares about. For example, the distinction between tactical AI versus strategic AI, and automation vs autonomy.

• Tactical is real-time, edge-deployed, and narrow, which includes fire control, threat identification, and route optimization, often with no human in the loop needed. When you need a missile to hit a specific target, it needs to aim at it accurately and get it right.

• Strategic is long-horizon and advisory, including wargaming, logistics, and intelligence synthesis, with decision cycles measured in days or weeks where human deliberation is built in.

• Automation is deterministic and rule-based. Autonomy is learned, adaptive, and probabilistic. The ethical questions are different for each.

He pointed out how companies don’t have the right to a veto. Once a product is legal and a use case is government-sanctioned, the contractor’s consent becomes optional, which is something that defense contractors have always known. Anthropic stood on its red lines but got replaced.

Anshuman gave an example of how things work when it comes to scale. A single MQ-9 Reaper combat orbit requires over 150 people: 49 in mission control, 59 in launch and recovery, plus intelligence, weather, and maintenance staff. The U.S. Drone Dominance Program is targeting 300,000 platforms, which means they can’t scale human operators linearly. Smaller nations with limited military populations face the choice of adopting AI-augmented operations or accepting capability gaps against peer threats. Ukraine is the proof of concept.

On hypersonic missile defense, his argument was that this is a physics problem rather than a policy question. Detection-to-intercept windows run under 15 minutes, often under 5. Human cognitive and command latency are physically incompatible with that limitation. The “no AI in lethal decisions” position is isn’t viable when the decision needs to be made across 100+ data streams in under 5 minutes.

Anshuman believes that safe AI in a defense context doesn’t mean unused AI. It means interpretable and auditable AI. Can a commander understand why the system flagged a target? Can a post-incident review reconstruct the full decision chain? Those are engineering goals, and they are achievable today. AI should be used inside a framework that embeds rules of engagement, produces traceable decisions, and keeps humans accountable for outcomes.

In the end, leadership and administrations change, but the tech keeps getting built. Palantir, Anduril, Shield AI, and dozens of others exist because of this need, not just because of political reasons. Then the decision becomes about how AI is integrated and whether the frameworks governing that integration are designed well.

Theoretically, and if we had endless time to decide, AI governance would not be such a pressing issue. But the realities of the world we live in won’t wait for us to come up with a win-win-win situation, and practically, it’s important to understand our current limitations and keep working towards making the most ideal solution to protect as many people as possible. It was good to hear from Anshuman’s perspective as someone who knows in real-time what can or can’t be done at the moment.

Personally, I’m coming from the perspective of an individual citizen, rather than someone from the regulatory or defense side, so it was very interesting to hear what people in those industries and positions thought about when it came to AI at that level of usage. It’s easy for me to stand on a moral high ground when it comes to using AI for war purposes, as someone who doesn’t have to make these decisions on the daily or have my decisions affect a lot of people. Still, I wonder how our voices could be used more prominently when it comes to thinking about these questions to support those who are making those kinds of decisions.

We had a short roundtable discussion after the speakers’ sharings before we broke out into a mixer. There was lively chatter about the philosophical topics from the speakers and I got to meet some cool people like policy researchers and lawyers building with Claude. It gave me a few interesting ideas for future meetups as well.

For these kinds of meetups in the future, I will consider some additional formats, like preparing small question cards for the mixer afterwards, doing breakout groups led by moderators, or using the Socratic method to get a healthy debate going.

Thank you again to all that participated and the support for the meetup. I’ll be launching our next meetup this week about how artists are using Claude. Feel free to reach out if you have any ideas or requests.

Disclaimer: The views shared are those of the individual speakers and do not represent Anthropic or any other organization mentioned.