knight

26. India
On my way to MLE.

[knight @HereeInsane]

Age 26
Worked only as customer service representative after completing bachelors CS degree where I did not learn to code or anything, dumb decisions.
A year back started learning ML, maths, probability. Loved it but haven't been consistent.

[knight @HereeInsane]

@matt_elevenlabs 11B

[Emilchess @EmilSutovsky]

Time flies and @ChessbaseIndia celebrates its first jubilee - 10 years.

Ten years ago Sagar, who started it as a passionate player and enthusiast, got an agreement and blessing of @ChessBase - to use their name and brand, and to represent the company in India and SEA region.

Gukesh, Pragg and Arjun just started their journey.

Indian chess boom was nowhere in sight. The genius of Vishy, and successes of Harikrishna, Sasikiran, Ganguly, Negi, Humpy, Harika and others already put the country very high in the chess ierarchy - but something important was missing.

Here comes Sagar. Tireless promoter of the game. Genuinely interested in every aspect. Top pros, talented kids, open tournaments, female chess, school competitions - Sagar (and later his team) were everywhere. Countless videos, interviews, reports - all made with the love and respect for chess.

Comes Covid. And FIDE Online Olympiad 2020.

And the seeds of Sagar's work bloom overnight.

Real chess fever starts in India. Government, mainstream media and sponsors saw the interest - as a result we witness almost unprecedented rise of chess in any given country. It is so big, that the impact on global chess is massive.

Sagar and his team played a crucial role in this monumental shift.

Wishing to CBI to keep going strong to the benefit of every passionate chess player in this world. Bravo!

[Sridhar Vembu @svembu]

We understand the push for Zoho to go public. But let me state the reality: Arattai would very likely not have been built by a public company that faces quarter to quarter financial pressure. It was a "hopelessly foolish" project, and even our employees had expressed scepticism that Arattai would ever gain any traction.

We built it because we felt we need that kind of engineering capability in Bharat. We need a lot lot more of such capabilities in Bharat and we are on it.

We have some very ambitious, long range R&D projects going on in Zoho, including compilers, databases, OS, security, hardware, chip design, robotics (not to mention AI) and on and on. In addition, we have invested in many R&D intensive companies that we know won't make money soon.

Zoho is a kind of an industrial research lab that also makes money to fund itself. We essentially ignore short term profits, as long as we don't lose money. And we have a culture of founders and senior executives living frugally, like how good scientists and engineers in ISRO would live. To us that is the essence of Bharat. Japan operated that way when it was developing.

Imagine saying all that to Wall Street or Dalal Street!

[giyu_codes @giyu_codes]

🎉Officially hired🎉

I graduated Spring 2023 with zero internships, zero job offers, and a trash portfolio.

I spent two years in the trenches of running my own consulting & development agency... climbing an uphill battle with zero credentials and zero industry experience trying to work my way into better clients or into an actual job. I am GRATEFUL to have found some success towards the more recent months, but it was BRUTAL.

My girlfriend and I sacrificed EVERYTHING to reach this point.

No birthday celebrations, no holidays, didn't hang out with friends, penny pinching for groceries, asking friends for help financially, donating plasma, getting a job but finding out I won't be hired for 4 more months, getting ghosted by clients, doing work for clients and waiting for them to return (I miss you guys hmu), denying the previous job because I knew I could make more money than what they offered me for the responsibilities they were asking for, no concerts, no vacations, missing out on going to japan when we were supposed to go, didn't get married because of the lack of stability I could provide, and watching life just move past us and progress as I was stuck holding back my girlfriend during the best years of her life.

You get the point.

I just want you to know that you can make it too.

Whether you're working at a FAANG looking to transition into consulting, a recent graduate that is scared for their future within tech, or maybe you're just starting out in tech...

There is a path for you. I promise.

Just please put in the work.

Position: Senior Full Stack AI/ML Engineer (I hate these names man idk what to call it but I'll be revamping an SME of 80+ employees from the ground up with basic technology + AI integrations at every level within a small team)

Compensation: $230k + benefits

Thank you @jxnlco @yacineMTB @tunahorse21 and @verrsane for being a direct impact in my work ethic, skill development, and mindset for all things in and out of tech.

Thank you to the community as a whole.

Next story arc: the wagie cagie

[Lingo.dev @lingodotdev]

Goodbye StackOverflow 🥳

[tomato 🍅 @neembu_paani31]

I've joined this org on 12th may and it's been one week only and I've designed whole agentic framework from scratch and I'm only dev building it ☠️

[major tom @tailwiinder]

We rise, we fall,
We fuck it, we ball

[The Khel India @TheKhelIndia]

NEERAJ CHOPRA HAS BREACHED 90M 🔥

[Kpaxs @Kpaxs]

Kasparov on greatness.

[ℏεsam @Hesamation]

how to go from zero on GitHub to becoming a contributor at DeepMind, Google Summer of Code or other open source programs:

this is a good read if you’re hoping to land a job in this market.

[Michael Francis @MFrancis107]

@karpathy I still think the Voyager paper is one of the best AI papers. And I haven't seen the concept used in practice anywhere.

[Andrej Karpathy @karpathy]

We're missing (at least one) major paradigm for LLM learning. Not sure what to call it, possibly it has a name - system prompt learning?

Pretraining is for knowledge.
Finetuning (SL/RL) is for habitual behavior.

Both of these involve a change in parameters but a lot of human learning feels more like a change in system prompt. You encounter a problem, figure something out, then "remember" something in fairly explicit terms for the next time. E.g. "It seems when I encounter this and that kind of a problem, I should try this and that kind of an approach/solution". It feels more like taking notes for yourself, i.e. something like the "Memory" feature but not to store per-user random facts, but general/global problem solving knowledge and strategies. LLMs are quite literally like the guy in Memento, except we haven't given them their scratchpad yet. Note that this paradigm is also significantly more powerful and data efficient because a knowledge-guided "review" stage is a significantly higher dimensional feedback channel than a reward scaler.

I was prompted to jot down this shower of thoughts after reading through Claude's system prompt, which currently seems to be around 17,000 words, specifying not just basic behavior style/preferences (e.g. refuse various requests related to song lyrics) but also a large amount of general problem solving strategies, e.g.:

"If Claude is asked to count words, letters, and characters, it thinks step by step before answering the person. It explicitly counts the words, letters, or characters by assigning a number to each. It only answers the person once it has performed this explicit counting step."

This is to help Claude solve 'r' in strawberry etc. Imo this is not the kind of problem solving knowledge that should be baked into weights via Reinforcement Learning, or least not immediately/exclusively. And it certainly shouldn't come from human engineers writing system prompts by hand. It should come from System Prompt learning, which resembles RL in the setup, with the exception of the learning algorithm (edits vs gradient descent). A large section of the LLM system prompt could be written via system prompt learning, it would look a bit like the LLM writing a book for itself on how to solve problems. If this works it would be a new/powerful learning paradigm. With a lot of details left to figure out (how do the edits work? can/should you learn the edit system? how do you gradually move knowledge from the explicit system text to habitual weights, as humans seem to do? etc.).

[luffy @0xluffy]

i honestly wished that but if i hadn't sacrifice a lot of sleep in the last 2 years i wouldn't be where i am. i'd probably be the same as the rest of the singapore CS grads.

the reason why i am here today is because i compressed 4 years of experience into 2 (by working fulltime while studying CS fulltime)

[knight @HereeInsane]

Absolutely

[Romee Panchal @romitheguru]

This is still one of the best intro to neural networks and backpropagation algorithm.

[Romee Panchal @romitheguru]

All ML enthusiasts tapping into Deep Learning and NLP may have heard about the famous "Attention Is All You Need" paper. Implementing it from scratch is a great way to understand it. However, it has many prerequisites to understand it fully.

Below is the list of papers that you should read and implement before proceeding to the "Attention Is All You Need" paper.

Start with Basic Neural Networks:
- Implement a simple feed-forward neural network from scratch
- Focus on backpropagation and gradient descent
- Paper: "Gradient-Based Learning Applied to Document Recognition" (LeCun et al., 1998) - implement a simpler version

Sequence Models Basics:
- Build a basic RNN from scratch
- Implement LSTM cells
- Paper: "Long Short-Term Memory" (Hochreiter & Schmidhuber, 1997) - start with a basic version

Word Embeddings:
- Implement Word2Vec
- Paper: "Efficient Estimation of Word Representations in Vector Space" (Mikolov et al., 2013)
- Understanding embeddings is crucial for transformers

Basic Attention Mechanisms:
- Implement Bahdanau attention
- Paper: "Neural Machine Translation by Jointly Learning to Align and Translate" (Bahdanau et al., 2014)
- This is directly relevant to transformer attention

Sequence-to-Sequence Models:
- Build a basic seq2seq model with attention
- Paper: "Sequence to Sequence Learning with Neural Networks" (Sutskever et al., 2014)

Once you've implemented these, you'll have the foundation to understand:
- Attention mechanisms
- Positional encodings
- Multi-head attention
- Layer normalization
- Residual connections

These components all come together in the Transformer architecture.

[Z Fellows @zfellows]

Paul Graham: “finish what you start, because the best work happens in the final stage.”

[Z Fellows @zfellows]

“Determination is probably the biggest piece of success [as a founder]” - Mark Zuckerberg

[rohit @seatedro]

if you're tired, do it tired

[Steve Jobs Stories @SJobs_Stories]

"People talk a lot about brands. To me, brands are one simple thing, and that is trust. Your customers trust you. We don't spend a lot of time talking about 'the brand'. We spend a lot of time talking about how do we make the best products in the world." - Steve Jobs, 2006