Quite disagree with @devinamehra article in the mint today. Here’s a solid research-grounded counterpoint from CFA institute. rpc.cfainstitute.org/blogs/e…
In 2005 I interviewed for Lehman and was disappointed that I didn’t get the job
In 2005, the highest expected-value career move a finance grad could make was taking the Lehman Brothers offer.
A hedge fund manager on why the smartest people make the number one career mistake:
Sinan Xin joined Lehman's tech M&A group out of Penn in '06 — now runs a global EM tech long/short fund.
Sinan Xin explains:
"The number one mistake people make at this stage is going somewhere for what seems like a rational economic outcome."
"The reason people do that is because these are intelligent people — making what seems to be a very reasonable and attractive economic choice."
"But five, ten years down the line, you're surrounded by people who made the same choice you did — for a different and potentially better reason."
"That's when you find you're not as competitive. You're not as passionate about generating alpha on a large cap US portfolio. You're not the expert on squeezing a millionth of a millisecond out of a trade."
"Making a decision based on passion rather than pure expected value is more alpha down the line — because the fidelity in calculating those expected values is quite low."
"In 2005, the expected value of taking a job offer from Lehman Brothers was quite high. The outcomes were different."
"People place too much faith in their own ability to assign correct probabilities to uncertain outcomes."
Hello @ClaudeDevs please fix this asap github.com/anthropics/claude… - unable to use Claude because of this and its terror
Shan Shah retweeted
I am likely biased, but I thoroughly enjoyed this podcast!
open.spotify.com/episode/6MD…
Dakshana Foundation - a unique non-profit built by a top value investor and investing in the RoI ... of people! #podcasthon special open.spotify.com/episode/6MD… #nonprofit @DakshanaIndia @MohnishPabrai
“If working with a model is like replicating a skilled employee that’s been working with you for six months rather than … the first hour of their job, … a huge unlock in AI capabilities.” - @dwarkesh_sp This is exactly what I’m building in the investing domain! Join us!
Shan Shah retweeted
We’ve launched The Fortnight — our new LinkedIn newsletter.
Edition one dives into Speed & Edge.
If this is your kind of reading, subscribe and stay with us.
linkedin.com/feed/update/urn…
Timely! The full story of cloudflare from the very beginning!
A deep dive into @Cloudflare
open.spotify.com/episode/39o…
Cloudflare’s journey is the story of how a spam-tracking side project among three founders blossomed into the invisible backbone of the internet, accelerating and protecting 20% of all web traffic worldwide. From its unlikely beginnings in law school to its game-changing freemium model, Cloudflare redefined internet infrastructure—fusing speed, security, and radical transparency into a global platform. Today, they’re not just powering websites, but shaping the future of edge computing, zero trust security, and cloud connectivity. Dive in to discover how relentless innovation and a mission to “build a better internet” transformed Cloudflare into a critical force behind our digital world—one whose best work happens so seamlessly, you might never even know it’s there
Transcript - empor.top/us/NET
Hey @JeepIndia , I _own_ one of your cars. Yet your team keeps spam-calling me to buy… a Jeep. 🤦♂️
Stop harassing your existing customers and clean up your lazy call lists. Spam isn’t marketing it’s brand damage. Are you proud of this and you think this is how you’ll succeed?
Shan Shah retweeted
The story behind the @aflac duck and more! empor.top/us/AFL.html a fascinating read.
pivotal.substack.com/p/the-w…
"Unfortunately, there’s a natural human tendency to hedge our bets — to make design choices in our experiments such that a mediocre success is the most likely outcome. There’s also a tendency to only do experiments that you know in advance will work" 💯
stop the spam @ICICIBank @ICICIBank_Care ... 7 calls in about 3 hours during a working day, do you really think anyone is going to appreciate this? Horrible! What would you feel if someone did this to your executives?
searchengineland.com/meta-ad… And theverge.com/2023/10/31/2394… and so on mean that we’re now clearly entering a phase where “ads are for poor people”. So now how do you target the high spenders using ads?
Really saddened by the service and communication by @JeepIndia How can it take weeks upon weeks to fix my Meridian? For a premium vehicle I expect premium service but here even basic service isn’t provided. 😡😔
Shan Shah retweeted
Lately, tech twitters been depressing.
- bad job market
- layoffs
- engineers applying & not hearing back
- talk of things improving only in 2024
- even good brands like Dunzo struggling
Figured I'll share internal 📊 to cover the state of India tech hiring and add perspective
In which chatgpt finds a date inconsistency in an article published today in @livemint chat.openai.com/share/57f2f7… 👏 epaper.livemint.com/Home/MSh…
I wonder why high denomination notes (now 500) don’t come with an expiry date printed on them. By that date you’re required to deposit the note back. No hassles announcing note related changes. Reduces black money hoarding. All without surprises and guesswork. @DasShaktikanta
Shan Shah retweeted
GPT-4 "discovered" the same sorting algorithm as AlphaDev by removing "mov S P".
No RL needed. Can I publish this on nature?
here are the prompts I used chat.openai.com/share/95693d…
(excuse my idiotic typos, but gpt4 doesn't mind anyways)
Sorting algorithm underpins all critical softwares. DeepMind's AlphaDev speeds up sorting small sequences (3-5 items) by 70%.
Key takeaways:
* The main RL algorithm is based on AlphaZero that originally played Go, Chess & Shogi. Same idea applies to searching programs!
* Instead of optimizing over C code, they optimize assembly code instead. It's a deliberate choice to go low-level to squeeze out every instruction saving.
* The assembly code is then reverse-engineered by hand to C, and open-sourced in LLVM.
* Even though the representation network uses transformer, it is NOT a foundation model. The whole pipeline only works on sorting, and has to be re-trained for other tasks like hashing.
