@TrietVo2002

data pipelines @google search. ex economic fraud @roblox , cs+math @nyuniversity competitive in: chess, poker, algo 🇻🇳🇺🇸

Joined November 2025
an LLM is only as smart as its user - that is, if the user is dumb, their LLM is gonna be dumb as well
1
13
the simplest way to overcome this is to act intelligent and/or don't restraint it with your limited knowledge
3
the general public doesn't have enough healthy skepticism on benchmaxx marketing
2
562
> create a human mimicking machine > doesn't talk to it like a human to maximize efficiency
1
568
what if the great filter is actually the gdp growth rate? that if you don't advance your civilization fast enough, you get smited by nature and the universe for being an ineffective species
1
397
on america 250, much has been written about her decline, but not her former dominance the only empire in human history to allow true immigration. the closest to a world government. pax americana and the fastest pace of technological development in history
177
if you're a hardware company and you're not in asia, are you actually a hardware company?
102
one thing has consistently predicted technical greatness across all domains and it's love for the game, yet almost no interviews directly measure it
1
52
"intelligent" investors betting that a technology will disrupt tech companies
52
treaty of versailles signed
39
there's a very simple solution to the AI spending problem and it is to use open source chinese models
1
56
yearning for the old era of RL when real scientists have the ball to reinvent superintelligence from scratch instead of trying to replicate human
48
crypto bulls and bears alike are missing the real market for crypto: facilitating gambling and illegal activities. this market is big and growing rapidly in the LLM era. not real world commerce. underground world commerce. build things criminals love.
1
1
62
in america the best outcome for a political party is to lose elections so you can squabble about what you would do rather than actually be in charge to do it
51
what happened to the tech industry of the 2010s? it was full of futurists, big dreamers and true scientists, where did all of them go
41
all the tech companies trying to build their own tooling around AI agents haven't learned the lessons every AI startups learn for the past 2 years: when the model gets better all your work become useless
1
1
35
labs oneshotted entire categories every other days and yet there's no shortage of dummies thinking they're special and signing up to be the next victim
31
the llm=agi discourse mistakes complexity for difficulty. some things can not be done by 99% of human (ex: solve imo math) but not complex, other things r complex but can be done by most with sufficient motivation (ex: write a 5th grade math textbook)
1
1
30
even if llm progress stalls, you still have a tool capable of most non-complex tasks and complexity can always be decomposed, so it's very useful with good eng culture and tooling
1
26
Triet Vo retweeted
我们 DeepSeek 又开始降价了。 事实上 我们降得太多了多到都不知道该怎么才好。 用户来求我 说 求求您梁文峰先生我们降太多了。我们用户不习惯一直降价,在你出现之前 我们总是被宰猪。 我说 不 不 不你们会大获全胜 —— 价格更低!你们会比以往任何时候都大赢特赢 —— tokens 更便宜!
65
186
32
3,099
174,129