Submissions from twitter.com/karpathy

		Software 1.0 easily automates what you can specify, 2.0 what you can verify (twitter.com/karpathy)
		1 point by speckx 16 hours ago \| past \| 1 comment
		Software 2.0 easily automates what you can verify (twitter.com/karpathy)
		1 point by cjbarber 1 day ago \| past \| discuss
		Andrej Karpathy on improved Tesla FSD performance under HW4 (twitter.com/karpathy)
		1 point by maxutility 4 days ago \| past \| 1 comment
		Andrej Karpathy: I am unreasonably excited about self-driving (twitter.com/karpathy)
		3 points by bilsbie 4 days ago \| past \| discuss
		Andrej Karpathy reviews latest wide-release Tesla FSD version (twitter.com/karpathy)
		11 points by Rover222 5 days ago \| past \| 1 comment
		Agency > Intelligence (Karpathy) (twitter.com/karpathy)
		1 point by andsoitis 20 days ago \| past \| 1 comment
		Karpathy on DeepSeek-OCR paper: Are pixels better inputs to LLMs than text? (twitter.com/karpathy)
		410 points by JnBrymn 27 days ago \| past \| 173 comments
		Karpathy on image only input to LLMs (twitter.com/karpathy)
		3 points by jonbraun 28 days ago \| past
		Andrej Karpathy on X: "My pleasure to come on Dwarkesh (twitter.com/karpathy)
		5 points by bilsbie 30 days ago \| past \| 1 comment
		Nanochat (twitter.com/karpathy)
		5 points by jbegley 35 days ago \| past \| 2 comments
		LLMs are mortally terrified of exceptions (twitter.com/karpathy)
		305 points by nought 39 days ago \| past \| 151 comments
		Karpathy's comments on the Sutton/Dwarkesh podcast (twitter.com/karpathy)
		10 points by cckolon 47 days ago \| past
		GPT-5 Pro is suited for solving hard problems (twitter.com/karpathy)
		4 points by vismit2000 71 days ago \| past \| 7 comments
		Continuing the journey of optimal LLM-assisted coding experience (twitter.com/karpathy)
		2 points by tosh 84 days ago \| past
		How to build a thriving open-source community by writing code like bacteria do (twitter.com/karpathy)
		4 points by tzury 4 months ago \| past \| 1 comment
		Bacterial Code (twitter.com/karpathy)
		4 points by twapi 4 months ago \| past \| 1 comment
		The race for LLM "cognitive core" (twitter.com/karpathy)
		8 points by mfiguiere 4 months ago \| past
		Karpathy: "context engineering" over "prompt engineering" (twitter.com/karpathy)
		27 points by Michelangelo11 4 months ago \| past \| 4 comments
		Most LLMs return the same answer in number guessing game (twitter.com/karpathy)
		6 points by samuel246 5 months ago \| past
		Aspects to video generation that may not be fully appreciated (Andrej Karpathy) (twitter.com/karpathy)
		16 points by Michelangelo11 5 months ago \| past \| 1 comment
		Andrej Karpathy: Opaque, Non-Scriptable UI-Heavy Products Risk Obsolescence (twitter.com/karpathy)
		3 points by nsoonhui 5 months ago \| past
		Direct Optimizability over Finite Indexing (twitter.com/karpathy)
		2 points by simonpure 5 months ago \| past
		An attempt to explain (current) ChatGPT versions (twitter.com/karpathy)
		3 points by tosh 5 months ago \| past
		Andrej Karpathy on X: system Prompt Learning? (twitter.com/karpathy)
		3 points by bilsbie 6 months ago \| past
		We're missing (at least one) major paradigm for LLM learning (twitter.com/karpathy)
		5 points by mfiguiere 6 months ago \| past
		"Chatting" with LLM feels like using an 80s computer terminal (twitter.com/karpathy)
		4 points by tosh 6 months ago \| past
		Rhythm in AI-Assisted Coding (twitter.com/karpathy)
		1 point by gmays 6 months ago \| past
		The primary audience of your thing (product, service, library,) is now an LLM (twitter.com/karpathy)
		3 points by tosh 7 months ago \| past \| 1 comment
		LLMs flip the script on technology diffusion (twitter.com/karpathy)
		1 point by sonabinu 7 months ago \| past
		Power to the people: How LLMs flip the script on technology diffusion (twitter.com/karpathy)
		1 point by tosh 7 months ago \| past
		More