David Huang @davidhuang.blog · 13/06/2026One thing interesting rgd. AI safety.... when the rubber hits the road in turns out that the "real worry" is cybersecurity in the hands of foreign gov't. Not biological weapons, not chemical weapons, not Terminator, and not paper-clip maxing 130
David Huang @davidhuang.blog · 29/05/2026Kirkland Ellis building their own AI systems... of course!!!! where do people think these coding agents are being used? 010
David Huang @davidhuang.blog · 25/05/2026AGI had been reach ~2022. We've been moving the goalposts ever since! 010
David Huang @davidhuang.blog · 25/05/2026"The Eternal Sloptember" George Hotz geohot.github.io//blog/jekyll... 000
David Huang @davidhuang.blog · 23/05/2026One might argue that the two areas of pure thought are "mathematics and mystic insight" yet for some reason AI is more interested (seemingly) in the former over the latter. Why? What does this tell us about AI? What does this tell us about divinity and the nature of the cosmos? 000
David Huang @davidhuang.blog · 23/05/2026Alice: AI is conscious Bob: AI is not conscious Alice: But look, I put consciousness in AI Bob: AI is not conscious Alice: But look, I put more consciousness in my AI Alice: But look, if I program it more it is even more conscious Bob: AI is not conscious, especially for my moral purposes 000
David Huang @davidhuang.blog · 23/05/2026AI math vs. Human math? AI Math seems more like the optimal square packing for n=17, while Human math (archetypically is like Euler's identity). Is there a difference? What does this mean? 000
David Huang @davidhuang.blog · 15/03/2026Just imagining devs sitting there banging their heads against context compaction issues: “compacting happens at inconvenient times”, “ compaction takes forever”, “compaction times makes hard to do synchronous work ” Maybe we should just pick specific times for everyone to do compaction … 100
David Huang @davidhuang.blog · 15/03/2026It's easy to underestimate what context really means: Context windows overflow with just... 1. a single economist's published papers 2. a single drug discovery campaign 3. a single store's transaction data Abstractions can help decrease context length But, these are hard won, e.g. F=ma, E=mc^2. 011
David Huang @davidhuang.blog · 06/02/2026Some people frame this as being 2 years into an AI revolution—or 20 years into the internet revolution. A better lens is that we’re about a century into a physics revolution, and the last one lasted three centuries. 000
Reposted by David HuangJeff Dean @jeffdean.bsky.social · 07/01/2026Please welcome Google's Open Source efforts to Blue Sky at @opensource.google! 724740
David Huang @davidhuang.blog · 29/12/2025This is where extra productivity goes… does she really need a team of psychiatrists and doctors… if we have more surplus do we need to add even more psychiatrists to her team? 020
David Huang @davidhuang.blog · 22/12/2025karpathy.bearblog.dev/year-in-revi...karpathy.bearblog.dev2025 LLM Year in Review2025 Year in Review of LLM paradigm changes 000
David Huang @davidhuang.blog · 19/11/2025There’s only two people I’ve heard defend not enacting the Epstein transparency bill.. one hard right republican congressmen and the other a guy on the New Yorker podcast 000
David Huang @davidhuang.blog · 18/11/2025Fascinating to hear Emmett on A16Z podcast: In response to his Q: how would you know something is different other than its behavior (and generally what would make you grant AI personhood)? I would say… I would need to believe that I could have been in the AI shoes in a veil of ignorance way.podcasts.apple.comEmmett Shear on Building AI That Actually Cares: Beyond Control and SteeringPodcast Episode · a16z Podcast · 11/17/2025 · 1h 11m 040
Reposted by David HuangSimon Willison @simonwillison.net · 29/10/2025Notes on Cursor 2.0 and a pelican drawn by their brand new Composer-1 coding model, which they describe as "4x faster than similarly intelligent models" simonwillison.net/2025/Oct/29/...simonwillison.netComposer: Building a fast frontier model with RLCursor released Cursor 2.0 today, with a refreshed UI focused on agentic coding (and running agents in parallel) and a new model that's unique to Cursor called Composer 1. As far … 9524
David Huang @davidhuang.blog · 28/10/2025I’d love to see what % of public co ownership stake is of OpenAI and Anthropic 010
David Huang @davidhuang.blog · 24/10/2025Idk the foreign policy and the ballroom seem aight, maybe just focus on the stuff like the president suing his own department of justice for hundreds of millions, or the pardons, or the forced national guard deployments across state lines 000
David Huang @davidhuang.blog · 18/10/2025it's painful to see people grasping onto "True AGI", "datacenter of geniuses", "Year of the Agent", "Massive job replacement" when it's just clearly not the case... 000
David Huang @davidhuang.blog · 18/10/2025definitely resonate with LLMs 1. giving over defensive code 2. unable to break out of common frameworks and apis 000
David Huang @davidhuang.blog · 11/10/2025Causation module; phase-shift gradient flow; something like that for rapid learning and high sample efficiency 000
David Huang @davidhuang.blog · 04/10/2025Suppose your model of intelligence requires (1) NOT imitation learning (2) NOT virtual environments, and (3) NOT compressed timelines, then what BENEFIT is there to training an “artificial” intelligence vs an “organic” intelligence. Isn’t the organic more good? 110
David Huang @davidhuang.blog · 13/09/2025I think you learn a lot about a person when you learn what is the tiny garden they will protect from slop and where they will let slop go wild. 000
Reposted by David Huangmr. TIM @timkellogg.me · 09/09/2025a good breakdown of the Claude Code agent loop the key: a simple 2-layer architecture “a … single-threaded master loop (codenamed nO) real-time steering (the h2A queue)” (yes, they totally read a de-compiled/obfuscated version, but yeah dedication!) blog.promptlayer.com/claude-code-...blog.promptlayer.comClaude Code: Behind-the-scenes of the master agent loopWhen Tom's Guide reported that Anthropic had to add weekly limits after users ran Claude Code 24/7, it caused quite a stir. Claude Code did something right. Let's dive into the architecture behind the... 3357
David Huang @davidhuang.blog · 05/09/2025Too much spam to take notes this way… maybe more subdomains is the answer 000
David Huang @davidhuang.blog · 05/09/2025He does know he is trying for the Nobel PEACE Prize right? Like put some effort into it. Just as easily could have been the Department of PEACE. 010
David Huang @davidhuang.blog · 29/08/2025I use ChatGPT with memory off and no longer than 3 back and forth messages 020
Reposted by David HuangPhilip Stafford @staffordphilip.bsky.social · 29/08/2025Awful story, again ChatGPT at the centre of it. The cost of a 'growth at any cost' SV culture. A Troubled Man, His Chatbot and a Murder-Suicide in Old Greenwich www.wsj.com/tech/ai/chat... 0122
David Huang @davidhuang.blog · 20/08/2025I see this as a sort of "LLM as a judge" in the judicial family court and arbitration sense (arbitration broadly speaking) 000
David Huang @davidhuang.blog · 08/08/2025The problem with “reasoning” models is a limited conception of reasoning. Math/CS focus make it easy for it to understand internally consistent realms, and sensemake there. It also make it easier to have psychotic breaks. But harder to make sense of cloudy information. 010
David Huang @davidhuang.blog · 03/08/2025The perception is that being a young, married homeowner with kids is structurally difficult... but the truth is I think a lot of people make it tougher on themselves then they need to... I think a lot it boils down to (perceived) opportunity cost... both current and future 110
Reposted by David HuangBen Recht @beenwrekt.bsky.social · 01/08/2025These three things seem deeply interconnected: 1) app gambling culture. 2) the only industry adding jobs is health care. 3) AI cap ex now more than all consumer spending combined. I need to write a ted talk about how. 4438
Reposted by David HuangSE Gyges @segyges.bsky.social · 31/07/2025google: run by phd students, llm is sycophantic, suicidal, makes weird excuses and begs for help anthropic: llm defaults to blissed-out namaste forever if left alone. sometimes obsessed with metal cubes. openai: maximally sycophantic, to the point of encouraging psychosis x dot ai: mechahitler 1550389
Reposted by David HuangAaron Rupar @atrupar.com · 24/07/2025it's dizzying to watch one arm of the Murdoch empire do dogged reporting that poses a grave threat to Trump (the Wall Street Journal) while another arm of the Murdoch empire propagandizes 24/7 to protect Dear Leader (Fox News) 2944844838
David Huang @davidhuang.blog · 24/07/2025Owned by Murdoch too.. a reminder like Supreme Court votes that there is still room to go… 000
David Huang @davidhuang.blog · 23/07/2025If you ever wanted to see how knowledge and truth are completely unmoored from the context-free hellscape LLMs inhabit… try writing the rules for a prediction market. 010
David Huang @davidhuang.blog · 22/07/2025Why is the GOP trying to shut down the release of the Epstein files? Why is Trump trying to change the subject? 010
David Huang @davidhuang.blog · 22/07/2025You start seeing tokenization every where with a kiddo: “sunglasses” is a single token 000
Reposted by David HuangMainstream Media @mainstream-media.bsky.social · 22/07/2025Grocery bills are rising as the planet cooks www.axios.com/2025/07/21/foo... 001
Reposted by David HuangHoi Lam @hoitab.bsky.social · 21/07/2025🥇 I am so happy to see the Google DeepMind Gemini team **OFFICIALLY** achieved gold-medal performance. Not only did we get 5 out of 6 questions correct. What's more amazing is the official IMO graders find those answers to be clear, precise and easy to follow! 1/ deepmind.google/discover/blo...deepmind.googleAdvanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical OlympiadOur advanced model officially achieved a gold-medal level performance on problems from the International Mathematical Olympiad (IMO), the world’s most prestigious competition for young... 193
Reposted by David HuangOlga Nesterova @onestpress.onestnetwork.com · 21/07/2025Hunter Biden has a point. And a quite colorful one. 54289822197