crazy world

2025 was a pivotal year for me. i achieved things i once only dared to dream of.

among all the things i’ve been gifted with this year, i’d say the most valuable is the mindset shift.

in the beginning of 2025, i was more of a realistic person. if you had come up to me and said you wanted to build a billion-dollar company in a year, i’d have said you’re crazy.

but i met some crazy people, and then i started to realize: those who are crazy enough about something are the ones who actually manage to achieve it.

so now, if you come up to me and say you want to build an ai data center in orbit, i’d ask where we should start.

there is no room for doubt in this world. if one has enough conviction and the means to do it, the last thing everyone should do is stand in their way.

breaking limits

in 2025, i broke many of my limits, achieving results in weeks or days that usually take months.

i built and designed a robot from scratch, created a gym for it, and trained an rl policy in two weeks.

this led to shahvir (my friend and past co-founder) and me raising $500k from south park commons and fr8 just four months into the journey.

i also received my first patent and had several articles on my creations featured on hackaday and hackster.io, both are platforms i’ve followed since secondary school.

i got into places like fr8, the residency, pluto, and south park commons.

i even managed to secure more than 10 sponsorships and monetized my youtube channel on the side.

most importantly, i never stop moving

before christmas, my friend arnie and i managed to secure a humanoid and became the first team in the world to reproduce twist2 along with its data collection pipeline. we did it in a single week.

bear in mind, arnie had never worked with a humanoid before—even an eth professor doubted him—but he managed to secure an $80k robot and space for us to work on it for a week. it’s this kind of mentality that inspires me.

as of now, i’m reading papers and designing a special vla architecture i’ve been thinking of for ultra-low-cost robots, while learning world models with a few friends.

you never stop once you’re in a true flow state.

the limit to what you can do is only mental

especially in this age of connectivity and ai, excuses are indeed just excuses. if you want to do something, now is the time to do it, no matter how crazy it is.

i did all of this while managing university, by the way. i’ll find out if i graduate in a week.

but everything comes with a cost, and i wouldn’t say 2025 was without one.

i left human computer lab this year after we raised our funding.

i stopped caring about my health; most of the weight i lost last year has now been regained.

i spent less and less time with my family.

at the end of the day, i still have a lot to learn and improve.

next year

next year, i’ll hopefully be moving back and forth between sf, shenzhen, and vietnam, spending time with colleagues, sponsors, friends, and robots.

i will also begin my new journey at livekit robotics, which i’m incredibly excited about.

my goal is simple:

write/build sota articles and projects every month.
build my home lab for "ultra-tinkering."
have fun.

right now, my focus is on world models, multi-modal llms, and building robots from scratch. my optimization vector is focused on latency, reliability, and generalization.

and my crazy thought for next year is that i’ll be able to publish three sota papers on robot learning and win best paper at neurips + icra.

we’ll see.

the most important part of this article

this year couldn’t have been achieved without my friends and family.

i’ve been blessed to meet and get to know some of the most incredible people on earth this year, some who might very well end up on mars and be the most incredible people there.

i can’t express my gratitude enough.

i truly love you all.