Jeremy Soo

Curve Labs at Open Robotics Al Forum SG. Co-hosted w/ OpenMind, Al Tinkerers. Joint by friends from Menlo, Tesseract, EDB, IMDA, NRP, Hyundai Cradle, DCG, Pantera, TGB, Auki, Monad, Virtuals, Kabam, and more at Glass Dome. A Token2049 side event supported by Gaib, Aeon, Peaq 🫶

reject cynicism dressed as pragmatism

@0x_ZHUANG Absolutely, can link you up with many!

@jeremysoojk wow, thats amazing to hear! I def need to meet more high agency, risk takers in Singapore would love to catch up

Was asked today on how the start-up space in Singapore looks among youth 🇸🇬 From my lens, peers from top universities rarely go after the entrepreneurial route. The default path is still clear: internships → grad role → climb. Safe & predictable. and I do think Singaporeans are far more risk-averse than they should be. A few reasons. Firstly, National Service. Two years from 18–20 sounds small on paper, but it removes a critical window, the most “risk-free” years of your life. By the time men graduate at 24–26, the clock already feels like it’s ticking. Less room to experiment, more pressure to settle down and to get it right immediately. Secondly, cost of living. Singapore is expensive. Every year a startup isn’t profitable, it’s not just “experience”, it’s real financial pressure. Rent, lifestyle, expectations. Combine that with graduating later, and suddenly you’re expected to think about stability, housing, and long-term commitments right out of school. Risk becomes a luxury. Thirdly, culture. Singapore prides itself as a talent hub. However, if we’re honest, it’s a hub for world-class employees, not risk-takers. Our system trains you to follow rules, optimise for grades, and win in structured environments. Not to break things and build from zero. Even at a personal level, I remember making my first 10k months at 19 running a sneaker shop. Instead of pride, it was met with skepticism. “Not stable.” “Not a real path.” The expectation was always: go corporate, climb the ladder, earn your stripes. That mindset is deeply ingrained. And it’s a shame, because Singapore has insane talent density. Smart, driven, resourceful people. Yet so many end up as middle/upper management in overseas giants instead of building their own. Of course, there are exceptions. The hungry few who take the leap and make it work. I can’t help but wonder, if even a fraction more took that risk, Singapore would not just be a “talent” hub, it would become the epicenter of innovation in Asia. The ingredients are already here: dense talent, strong infrastructure, smart capital, global connectivity. If that mindset ever shifts, Singapore become more than just a place that produces talent. It starts becoming the place where Asia’s next generation of iconic companies are born.

hello from singapore at 1:30am @fdotinc

@fdotinc GM

gm chat

@Josephayinde64 @fdotinc Time to build 🫡

@fdotinc where are my canopy folks at?

see you in @fdotinc canopy building empathic ai @curve_labs starting with tooling & infra for humanlike memory, persona control, emotional states also building out apac robotics ecosystem long term these will merge into one future filled with feeling machines e.g. baymax

@yoeven @FarzaTV YOOOO legends

Full circle moment seeing @FarzaTV on my first day at YC! I first came to SF for buildspace s5 with just a small idea for JigsawStack. Now we're both in the same YC batch figuring shit out!

@yoeven Yessir yessir 🫡

if you're cracked, you'll be able to reply

@yacineMTB It's always been long horizon consistency 😔

I have found the limits of gpt 5.4

"Just do things" at its extreme is identical to "act always on instinct" i.e. maximize P(get what I want/need now | all else), which, since humans dominate precisely because we don't do that unlike all other animals, is almost guaranteed ineffective long term.

@agrimsingh @ivanleomk @SherryYanJiang @cursor_ai @nickwm @fr4nnyp4ck LET'S GOOOO

you asked for it, you get it @SherryYanJiang and i are finally bringing cafe @cursor_ai singapore aka cursor kopitiam to town @nickwm and @fr4nnyp4ck will be flying in to drink some coffee and hang out spend the evening hacking or just catching up with other cursor-pilled folks registration linked below:

I've hit the limit. I'm re-writing my entire data pipeline by hand. I can't explain it, whether it's architecture choice, or tiny nuanced latency trade-offs, or just global coherence. The models do not feel smart at all when you already know how you want to solve a problem. Slop.

@Suhail So supply is short. And they sold more compute than they can deliver. A tale as old as time.

The run on inference capacity is coming. You have been warned.

1T 👀 Kimi 👀 K3 👀 Please

Coding agents work, just as you on a night out may conclude any hard drink works. Five part shots, one part mixer. Carelessly shaken. It works. But in your drunken stupor, you missed the part where some guy pissed in your drink for no reason at all. But sure, who cares. It works.

@lekt8_ Next event name 👀

funbrowse

@signulll The lesson here is really that tech stack lock-in during pre-training is real. Same reason React keeps winning. CLI is more native. Basic API endpoints are more native.

absolutely nuts that it’s been like only a year since mcp has been mainstream & it is already dying.

MD to database transition for agents will happen for the exact reasons humans moved off (counter to popular belief, extremely scalable) paper in drawers.

@Noahpinion @signulll Finite processing, infinite information

@signulll Retard question here: Why can't they just put it all in the context window?

no one has been able to solve ai memory yet. it’s brittle, it’s fragmented, & often times less helpful than not using memory. it’s an incredibly fascinating problem, way more of an art than a science at this point.