HFT Quant

Result of 1.5 years of hard work and study. Still not there, but slowly getting there. Out of sample simulation, subsecond bid/ask tick data, trading costs included. #MachineLearning #ML #AI #algotrading #marketmaking #FX #forex #forextrading

@karpathy self improving AGI incoming...

Thank you Jensen and NVIDIA! She’s a real beauty! I was told I’d be getting a secret gift, with a hint that it requires 20 amps. (So I knew it had to be good). She’ll make for a beautiful, spacious home for my Dobby the House Elf claw, among lots of other tinkering, thank you!!

USA has ChatGPT USA has Grok USA has Claude USA has Gemini USA has Llama USA has Copilot China has DeepSeek China has Qwen China has Ernie China has GLM China has Kimi What does your country have?

@Bitcoin_Teddy unethical?? why?? there's a game where players are trying to take each other's money. These guys put lots of effort to build a system to take the other players' money. Kudos to them. It's not like they are ripping of lonely grandmas with a romance scam.

This is just one of countless unethical ways money is made online. This video shows a poker bot farm. Multiple bots sit at the same table and share their cards in real time. Because they know each other’s cards, they never bluff or trap each other. They only bet aggressively when the human is statistically behind, and fold otherwise. A single bot farm like this can make more than the house!!

so... is the fly conscious? 🧠🤔

@elonmusk Can 𝕏 Money be frozen? If yes, than GTFO

𝕏 Money early public access will launch next month

@karpathy Andrej, how about a SETI@home style distributed autoresearcher powered by community CPU/GPU compute, for everyone’s benefit? :)

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes, today I measured that the leaderboard's "Time to GPT-2" drops from 2.02 hours to 1.80 hours (~11% improvement), this will be the new leaderboard entry. So yes, these are real improvements and they make an actual difference. I am mildly surprised that my very first naive attempt already worked this well on top of what I thought was already a fairly manually well-tuned project. This is a first for me because I am very used to doing the iterative optimization of neural network training manually. You come up with ideas, you implement them, you check if they work (better validation loss), you come up with new ideas based on that, you read some papers for inspiration, etc etc. This is the bread and butter of what I do daily for 2 decades. Seeing the agent do this entire workflow end-to-end and all by itself as it worked through approx. 700 changes autonomously is wild. It really looked at the sequence of results of experiments and used that to plan the next ones. It's not novel, ground-breaking "research" (yet), but all the adjustments are "real", I didn't find them manually previously, and they stack up and actually improved nanochat. Among the bigger things e.g.: - It noticed an oversight that my parameterless QKnorm didn't have a scaler multiplier attached, so my attention was too diffuse. The agent found multipliers to sharpen it, pointing to future work. - It found that the Value Embeddings really like regularization and I wasn't applying any (oops). - It found that my banded attention was too conservative (i forgot to tune it). - It found that AdamW betas were all messed up. - It tuned the weight decay schedule. - It tuned the network initialization. This is on top of all the tuning I've already done over a good amount of time. The exact commit is here, from this "round 1" of autoresearch. I am going to kick off "round 2", and in parallel I am looking at how multiple agents can collaborate to unlock parallelism. github.com/karpathy/nanoc… All LLM frontier labs will do this. It's the final boss battle. It's a lot more complex at scale of course - you don't just have a single train. py file to tune. But doing it is "just engineering" and it's going to work. You spin up a swarm of agents, you have them collaborate to tune smaller models, you promote the most promising ideas to increasingly larger scales, and humans (optionally) contribute on the edges. And more generally, *any* metric you care about that is reasonably efficient to evaluate (or that has more efficient proxy metrics such as training a smaller network) can be autoresearched by an agent swarm. It's worth thinking about whether your problem falls into this bucket too.

@Bitcoin_Teddy plot twist: US attacked Iran because Iran has the cheapest cost of electricity $0.01 per kW, so that they could take over and mine Bitcoin....

Bitcoin has never fallen below its electrical cost. Today that floor is $71,000.

@BrissyBitcoiner @UTXOTimes you can tell from the way he handles it, this is heavy AF, that's why it's real.

@UTXOTimes how do you verify they are real gold?

SELLING ALL MY GOLD HERE. BEEN A NICE RUN BUT ITS TIME TO MOVE TO #BITCOIN. NOT FINANCIAL ADVICE

@PrisonPlanet If I was selling underground bunkers, this would be the perfect viral thing to say to sell more bunkers... Who texted him exactly?

Owner of company that builds nuclear Armageddon-proof underground bunkers Ron Hubbard says two senior Trump administration cabinet members have become new customers. "One of them texted me yesterday, asking me: ‘When will my bunker be ready?" Slightly concerning.

@KaladinFree @kalinowski007 "Elected officials who can be held accountable" 🤡🤡🤡

@kalinowski007 With all due respect I don’t want people like you, or any company for that matter, to be the decision makers on things like this. Elected officials who can be held accountable should be the decision makers.

I resigned from OpenAI. I care deeply about the Robotics team and the work we built together. This wasn’t an easy call. AI has an important role in national security. But surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got. This was about principle, not people. I have deep respect for Sam and the team, and I’m proud of what we built together.

There isn’t enough Bitcoin for everyone.

@Vivek4real_ How many years this dude cried wolf now??? 👎

BREAKING: 🇺🇸 RAY DALIO JUST DROPPED A MASSIVE WARNING AGAINST BITCOIN: “BITCOIN DOES NOT HAVE PRIVACY. ANY TRANSACTION CAN BE MONITORED.” “CENTRAL BANKS DON’T WANT TO BUY BITCOIN.” “AND THERE ARE ALSO QUANTUM COMPUTING ISSUES.” “IT’S A RELATIVELY SMALL AND CONTROLLED MARKET.”

@lady_valor_07 bunch of obsolete cables, flashlight that does not work, foreign coins you can't spend, keys you can't use, 32MB MP3 player, dead batteries...

starting a junk drawer. What else do I need?

Calling LLM a "stochastic parrot" isn't insight from brilliant professors and sharp minds, it's the last gasp of intellectuals whose own intelligence is threatened... and getting demolished in all domains where you thought you were untouchable... #COPE AGI is coming 💀🔥#AI #LLMs

@saylor 🫡 yes sir!

I'm buying bitcoin right now. Are you?

@wholyv when you do AI/ML in python ~97% of GPU/CPU time is already spent in well optimized c/c++ code. If you wrote all in c++, you would get like +3% improvement.

What if C++ was used for machine learning instead of Python. given C++ evolved to allow that. I wonder how faster, our AI models would be these days.

That's hot👍 Someone hire this guy!

@MichaelOnBTC of course! 💯

Would you?

@stithprajnya lol. Dude I have no control over what you read or reply to... you're wasting your own time. Don't blame me!

@QuantRob What is cnn? Stop wasting my time, markets will humble you.

hot take: any quant who is not using LLMs for research in major way will be replaced within a year. Thoughts?

@AbhiCodes15 can't have 4 chefs in the same kitchen... this startup would fold within a month...

If these four are building an AI startup together, who should be the CEO?