ABOUT — PERSON AT WORK
Most ideas are wrong.
What I care about is how they get killed.
I'm Elias. I've coached physics olympiads, read a regulated industry from scratch in four months, and built alpha signals in a quant competition where I was also the one saying no.
THINGS THAT DON'T CHANGE
2026The cost of no55 of the 60 signals were mine. Every one I cut took money out of someone's pocket.
In the 2026 WorldQuant IQC, 55 of the 60 signals in our final portfolio were mine. Five were my teammates'. I rejected about twenty more.
I like running the project. I'm not going to dress that up. But it had a price: we signed an agreement at the start splitting prize money by submission count. So every signal I cut took money out of someone's pocket.
One teammate was submitting machine-ensembled signals in volume. I judged them overfit. To actually stop it I had to understand why he was doing it — the leaderboard was live, he wanted to contribute volume, and he genuinely believed enough stacking would eventually produce something. Only the third of those is an argument you can have.
We finished 1st of 373 Taiwan teams and 32nd globally. That only proves I wasn't wrong enough to lose.
2021—25The other side of noFour years of IPhO. Contempt is efficient, and that's the problem.
Before that I spent four years teaching olympiad physics. Two IPhO golds, one silver.
What I actually learned there: a student can only stay calm in front of a problem he's never seen if he believes that saying something stupid is safe.
Contempt is efficient. It lets you dispose of an idea without understanding it first. The cost is that you occasionally kill a good one — and more often, you kill that person's willingness to speak up next time.
2025—26Not knowingFour months to read industrial gas detection from zero, then hand it to someone else.
At Everrist I had four months to learn industrial gas detection from zero — sensor principles, Ex d / Ex ia certification, communication protocols — and turn it into something a person could make a decision with. Then I had to get a Korean R&D team to adopt a compound-selection tool covering 840+ substances.
Both of those start the same way: admitting you don't know yet.
2016—24The long way roundSeven and a half years for one degree. I'd rather say it plainly than round it down.
My EE degree at NTU took seven and a half years. I stopped in the middle to start a company. I'm doing a CS master's at UIUC now, online, alongside work.
I'd rather say that plainly than round the dates down.
THE RECORD
Alongside work, on purpose.
1st of 373 Taiwan teams, 32nd globally.
Industrial gas detection, from zero.
Three IPhO golds, one silver. Team of three people.
Seven and a half years. I stopped to run a company.