🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

i am really good at short form writing and i'm able to string together the kind of words i love, but the moment i try to expand an idea, it feels like the coherence is falling off and i need to put a lot more energy to achieve the same feeling. any tips on how to get around this?

shocking that you can string together a bunch of words and win the nobel prize, build a billion dollar company or like get to mars. you just need to predict the right next token

ok so the haber bosch process essentially removed the natural limit on how many people the earth could feed by converting n2 into nh3 (ammonia), leading to synthetic fertilizers, which led to increased crop yields, which in turn led to more food production. damn

i wish we all had the time to explore our curiosities instead of holding them back on the side so that we can achieve our real goals first

a lot of smart people early on try to rationalise everything. putting numbers next to a thing makes them feel like they now understand it. this makes it very hard for them when they want to solve valuable problems. the really valuable problems are too vague, abstract, and hard in

when you are embedded in any system, it is easy to start optimising for the metrics that system values rather than stepping back to ask what actually matters. those local metrics often reward activities that create minimal real world impact and crowd out time that could be spent

the cult of done manifesto

can't believe this app is free

day in ml grind 34 9 hours >spent most of the day writing pytorch with some math >performed style transfer using variational autoencoders >jordan normal form of matrices, solved a few linear algebra problems. >gained a little intuition for rlhf on language models

when we use SAEs to understand models, we’re trading off fidelity for insight. we make the model’s thoughts legible by compressing them into human friendly pieces, but we inevitably leave behind some of the richness that made those thoughts possible in the first place. i wonder

the best ideas often emerge from sitting with uncertainty, letting contradictory thoughts bump against each other, exploring dead ends that inform the eventual path forward. it's inherently inefficient-looking work. lots of staring out windows, scribbling and crossing out, having

when new inputs become available (like ai capabilities), there's often a window where applications can capture outsized value by being among the first to effectively harness those inputs for real world problems. it's all better ways to utilize energy at the end of the day

value used to accrue to whoever controlled the atoms: land, factories, oil. today it accrues to whoever conquers the bits: attention, data, mindshare.

I've been getting wayy too many dms lately regd how to dj. Tldr :

unfortunately, what the world lacks most is not technical ability but ideas and individuals with the conviction to nurture them into something real

your ability to ship features at a faster pace might give you a 1 month lead, but if those features don't add any tangible user value, then speed is a curse. it will make it seem like you're making progress while you're actually heading toward a slow death