Arunim Agarwal
1 followers · 8 following · 269 views
on the atlas — 82
- Dario Amodei — We Must Pace the Frontier31 savers
- AI Safety in China: A Primer - by Jasmine Li - The J-Space2 savers
- How Should the US Prepare for Increasingly Automated AI R&D? | IFP3 savers
- On Doubling9 savers
- MECE principle2 savers
- Are we existentially threatened by the type of AI misalignment seen in the OpenAI Hugging Face attack?1 savers
- Rule of 722 savers
- Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident - METR20 savers
- What just happened? Pragmatism and Pessimization — LessWrong7 savers
- The Nvidia-sized hole in US GDP statistics | Epoch AI2 savers
- Computational Complexity: Impagliazzo's Five Worlds3 savers
- Statecraft for the AI Transition - Center for Technology & Statecraft1 savers
- Challenges of governing continually learning AI – Runtime Verification1 savers
- Why You Should Almost Never Use AI to Write Anything Substantive5 savers
- Will parallelization limits delay an intelligence explosion? | Epoch AI1 savers
- GDP and prosperity across technology paths1 savers
- A Safe Path to Open Weights - Thinking Machines Lab9 savers
- Mini Blog Post 3: Become a person who Actually Does Things — Neel Nanda45 savers
- Investigating three real-world incidents in our cybersecurity evaluations \ Anthropic4 savers
- “Radiance: A Novel”, by Carter Scholz, Gregory Benford, Hugh Gusterson, Samuel T. Cohen, Curtis LeMay · Gwern.net2 savers
- OpenAI and Hugging Face partner to address security incident during model evaluation | OpenAI14 savers
- What we’d like to fund — Paradigm 35 savers
- Simple advice for academic publishing - Marginal REVOLUTION5 savers
- Folk theorem (game theory)3 savers
- Advice That Actually Worked For Me - Nabeel S. Qureshi7 savers
- Quote Receipts - Teddy Warner2 savers
- Risk-Averse AIs4 savers
- The Future Worth Building Is Human - Thinking Machines Lab23 savers
- AI 2040: Plan A9 savers
- Selective Optimism: a critique of AI 2040 - by Richard Ngo6 savers
- A Long Sequence of Small, Correct Decisions3 savers
- No one can teach you to have conviction | benkuhn.net39 savers
- Rest in motion24 savers
- A reading list for generalists — LessWrong18 savers
- What Makes Art Great? - Nabeel S. Qureshi7 savers
- Rented Virtue - by Will Manidis - Minutes3 savers
- The high-return activity of raising others' aspirations - Marginal REVOLUTION18 savers
- First, Make Me Care, by Gwern · Gwern.net11 savers
- John Marshall1 savers
- More compute, more capability: Why AI agent evaluations need to account for test-time compute | AISI Work1 savers
- The Pilgrims1 savers
- The date of AI Takeover is not the day the AI takes over — LessWrong1 savers
- [REPOST] Epistemic Learned Helplessness | Slate Star Codex13 savers
- On Seeing Through and Unseeing: The Hacker Mindset · Gwern.net18 savers
- Who should the U.S. Talk to in China on AI?1 savers
- Reinforcement learning towards broadly and persistently beneficial models6 savers
- Toward an O*NET for AI R&D2 savers
- Bushido1 savers
- Cognition Café2 savers
- Inside the whirlwind 24 hours that led the White House to slap export controls on Anthropic - POLITICO2 savers
- Statement on the US government directive to suspend access to Fable 5 and Mythos 5 \ Anthropic9 savers
- Dario Amodei — Policy on the AI Exponential16 savers
- Guardian Angels: LLM Personalization for Productivity and Security · Gwern.net25 savers
- We Should Think More Critically About Compute Verification | Sudarsh Kunnavakkam2 savers
- Built to benefit everyone: our plan | OpenAI7 savers
- Useful Vices for Wicked Problems7 savers
- How to Secure the DNA Supply Chain | IFP1 savers
- Javier Milei: Argentina invites AI to free itself1 savers
- The fall of the theorem economy - David Bessis9 savers
- The importance of AI character3 savers
- A playbook for field strategy - by Dewi Erwan5 savers
- Quality Technology Services1 savers
- People's deeply held beliefs are surprisingly surface-level1 savers
- The Rogue Replication Threat Model - METR1 savers
- Casey_at_the_Bat1 savers
- How Should We Implement an AI Pause? | Sudarsh Kunnavakkam1 savers
- AI and TikTok are giving brain rot. Could flash cards reverse it? | The San Francisco Standard1 savers
- What if the AI line just keeps going up? - by Jesse2 savers
- How To Be Successful96 savers
- Staring into the abyss as a core life skill85 savers
- How I've run major projects | benkuhn.net52 savers
- Do Thing, Do One Thing49 savers
- Encyclical Letter of His Holiness Leo XIV Magnifica Humanitas (15 May 2026)46 savers
- Meditations On Moloch | Slate Star Codex32 savers
- Impact, agency, and taste | benkuhn.net30 savers
- The Top Idea in Your Mind14 savers
- The advice I would give on a mentorship call10 savers
- The Goddess of Everything Else | Slate Star Codex8 savers
- Life is Short7 savers
- Antiqua et nova. Note on the Relationship Between Artificial Intelligence and Human Intelligence (28 January 2025)6 savers
- The "Messy Middle" - by Molly Kinder4 savers
- AI Deterrence by Betrayal4 savers
highlights — 17
The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks. This power must be scoped to the above four specific risks and there must be protective measures against political favoritism or arbitrary decisions.
Dario Amodei — Policy on the AI ExponentialIf this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.
Statement on the US government directive to suspend access to Fable 5 and Mythos 5 \ AnthropicThe wave will be slower than expected.
We Need Unhobbled Donors - by Saheb GulatiAn individual donor provides value to the field because they have a distinct theory of victory and risk appetite.
We Need Unhobbled Donors - by Saheb GulatiHe is not content with any one good or great deed, but always seeks to top himself; and furthermore, for all the worldly esteem he may attain, he, in all his greatness, knows that not even the esteem of history is the most important a man can attain, but that of God
Character of the Happy WarriorHis law and dearest friend is Reason; he owes all of his triumphs to Virtue. If he achieves high station, he does so honestly; if he cannot act honestly whilst in office, he would sooner quit. Because he is single-mindedly faithful, he does not seek his own advancement.
Character of the Happy WarriorAll fearsome challenges he transmutes, subduing what negative qualities they may have, and learning from what good they have to offer. The warrior is "skilful in self-knowledge" (like the philosophers of ancient Greece, living by the famous injunction to "know one's self") and understands that the true purpose of "suffering and distress" is to grow in compassion and "tenderness".
Character of the Happy WarriorThe Happy Warrior is a generous spirit, who, amidst, or, in spite of, the tasks of real life, hath done what pleased his innocent, "childish thought". His noble ideas and deeds are "an inward light" (not unlike the Quaker belief in an inner light) that, despite their inwardness, make the path before the warrior "always bright."
Character of the Happy Warriorin a large group, everyone needs to be autonomously making frequent, high-quality, local prioritization decisions, without needing a round-trip through me. To get there, they need to be ambiently aware of: what else is going on around them, so they can coordinate and update on new info quickly (“oh, we’re planning to kick off the next derisking run in three days, so I have to have my new RL environment ready and tested by then”) how their goal fits into the overall project, so they can make correct decisions about the details of their approach (“we’re trying to scale up as much as possible rig…
How I've run major projects | benkuhn.netRun a fast OODA loop OODA stands for “observe, orient, decide, act”—in other words, the process by which you update your plans and behavior based on new information.
How I've run major projects | benkuhn.netThe plan is important because whether or not we’re achieving the plan is the best way to figure out how well or badly things are going. Knowing how well or badly things are going is important because it tells me when to start asking for more support, cutting scope, escalating problems, and otherwise sounding more alarms. One of the most common megaproject failure modes is to not freak out soon enough, and having a concrete plan is the best antidote.
How I've run major projects | benkuhn.netA specific tool that I’ve found critical for staying oriented and updating quickly is a detailed plan for victory, i.e., a list of steps, as concrete as possible, that end with the goal being achieved.
How I've run major projects | benkuhn.netEven more importantly than freeing up time, clearing my schedule made sure the project was the top idea in my mind. If I don’t do that, it’s easy for me to let projects “go on autopilot,” where I keep them running but don’t proactively make time to think through things like whether we should change goals, add or drop priorities, or do other “non-obvious” things.
How I've run major projects | benkuhn.netif you look at 2025, Anthropic’s revenue was growing at an absurdly fast rate for that year, fast enough to reach 100% of global GDP by May 2030. You could reasonably look at that and say sure, their revenue growth is impressive, but the line can’t keep going up, it has to taper off eventually. No way they’ll hit that number by mid-2030. You’d probably have felt pretty confident in making that claim, I know a lot of people did. But what did the line do? It bent upwards. Revenue started growing at a cool 80x per year, instead of 10x per year.
What if the AI line just keeps going up? - by Jesseat its current rate of exponential growth, Anthropic ARR will hit 100% of current global GDP, i.e. around $126 trillion dollars, in December 2027.
What if the AI line just keeps going up? - by JesseEven if the White House ultimately decides to reject mandatory vetting of AI models, lobbyists increasingly believe such a regime could emerge at some point in the future — particularly when more advanced models emerge or if Democrats, who are generally more supportive of AI rules, take control of the government. To prepare for that possibility, some companies are weighing whether to participate in voluntary testing, according to the AI policy adviser. “You cannot avoid technological and political change,” the first lobbyist said. “So I suspect even if this effort ends in voluntary CAISI revie…
White House’s ‘lack of organization’ has AI lobbyists fretting - POLITICOThe administration’s recent messages to industry have complicated expectations that President Donald Trump would take a hands-off approach to overseeing AI. That assumption took a hit this spring with the emergence of Mythos, a new AI model developed by Anthropic with the potential to supercharge cybersecurity risks. Top Trump officials, including chief of staff Susie Wiles and National Cyber Director Sean Cairncross, moved quickly to address the potential threat — and soon began discussing a slate of possible actions, including an executive order that would require companies to get a green li…
White House’s ‘lack of organization’ has AI lobbyists fretting - POLITICO