all while watching my ports get blasted through

documenting this project is harder than building it playing bumper cars everyday

Do it!

iPhone users can relate

NEW: NYC comptroller Brad Lander whines after Elon Musk took back $80 million that the city was going to use on illegal immigrants, says the money just disappeared from their accounts. Great! The Trump admin issued a clawback after FEMA officials sent tens of millions of dollars to the city for illegal immigrants. "This is highway robbery. This is the federal government having already allocated, authorized, awarded, invoiced, and paid $80M to the city of New York, and then Elon Musk with no legal authority seizing it back."

This "pastor" needs to be investigated as soon as Kash Patel is confirmed.

🚨BREAKING: Alex Jones Predicts A Globalist Cyber Attack On Government Payment Systems To Be Blamed On Musk Watch/Share Alex Jones Cover What Happens Next! x.com/i/broadcasts/1zqKVjQzg…

I was the lead market maker at Jump. I'm no longer under NDA so I can say this, Typically when you see down moves during the weekends like this, it is pure market manipulation. It was one of the tactics I invented during the low volume moments. We would load up and pay attention to Crypto Twitter as a source of sentiment, with specific accounts organized in specific categories. One thing I used to like doing was purposely "nuking" a new 4H candle on BTC and/or ETH, which was an easy trap. "Smart" traders know we'll eventually revisit those wickless candles, so it was good bait for me to set with eager buyers stepping in and the candle continuing lower. They end up capitulating before we reverse. Bottom shorters get comfortable here too. Anyway, I can't give away too much but just know, Most of thos "bad" price action you're seeing is a group of "whales" sitting in a chatroom together and merely oil painting on charts. My alias in Jump was "Vincent van Gogh." And that isn't because I lost my mind [well - that too], But because I painted some of the best looking bear traps. I know an influenced painter copying my work when I see one. Stay safe. Retar Dio.

Most interesting reply to this thread thus far from @MichaelEMcNeil

I don't have too too much to add on top of this earlier post on V3 and I think it applies to R1 too (which is the more recent, thinking equivalent). I will say that Deep Learning has a legendary ravenous appetite for compute, like no other algorithm that has ever been developed in AI. You may not always be utilizing it fully but I would never bet against compute as the upper bound for achievable intelligence in the long run. Not just for an individual final training run, but also for the entire innovation / experimentation engine that silently underlies all the algorithmic innovations. Data has historically been seen as a separate category from compute, but even data is downstream of compute to a large extent - you can spend compute to create data. Tons of it. You've heard this called synthetic data generation, but less obviously, there is a very deep connection (equivalence even) between "synthetic data generation" and "reinforcement learning". In the trial-and-error learning process in RL, the "trial" is model generating (synthetic) data, which it then learns from based on the "error" (/reward). Conversely, when you generate synthetic data and then rank or filter it in any way, your filter is straight up equivalent to a 0-1 advantage function - congrats you're doing crappy RL. Last thought. Not sure if this is obvious. There are two major types of learning, in both children and in deep learning. There is 1) imitation learning (watch and repeat, i.e. pretraining, supervised finetuning), and 2) trial-and-error learning (reinforcement learning). My favorite simple example is AlphaGo - 1) is learning by imitating expert players, 2) is reinforcement learning to win the game. Almost every single shocking result of deep learning, and the source of all *magic* is always 2. 2 is significantly significantly more powerful. 2 is what surprises you. 2 is when the paddle learns to hit the ball behind the blocks in Breakout. 2 is when AlphaGo beats even Lee Sedol. And 2 is the "aha moment" when the DeepSeek (or o1 etc.) discovers that it works well to re-evaluate your assumptions, backtrack, try something else, etc. It's the solving strategies you see this model use in its chain of thought. It's how it goes back and forth thinking to itself. These thoughts are *emergent* (!!!) and this is actually seriously incredible, impressive and new (as in publicly available and documented etc.). The model could never learn this with 1 (by imitation), because the cognition of the model and the cognition of the human labeler is different. The human would never know to correctly annotate these kinds of solving strategies and what they should even look like. They have to be discovered during reinforcement learning as empirically and statistically useful towards a final outcome. (Last last thought/reference this time for real is that RL is powerful but RLHF is not. RLHF is not RL. I have a separate rant on that in an earlier tweet x.com/karpathy/status/182127…)

“Social-conservative Fiscal-liberal” are the worst type of Republicans, yet they’re ascendant in Congress. They’re tearing apart traditional family structures by wrecking our economy with big government spending.

spent hours tweaking the database structure beta testers needed maybe

waiting on an intro = lack of agency. just reach out

i give ross my thanks for joining crypto

kept the ui simple as possible. formatting was so tricky for user experience

.@arcprize quick project Need extra hands to help with r1 testing for ARC Prize. Paid Eng LLM work, python Know anyone? Dm or greg@arcprize

DeepSeekの推論AIモデル「DeepSeek-R1」を使ってコーディングできる(成果も表示)「DeepSeek R1 Code Generator」が、Hugging Face内のAIチャット「anychat」に実装されました。無料でお試しできます。