Links (4)

A savage place! as holy and enchanted; As e’er beneath a waning moon was haunted

Disclaimer: Just because I include a link does not mean I endorse it or am confident in its claims.

1. A body most amenable to experimentation: More fiction from Abhi. Models of Life is one of my all time favorite bio short stories, so I consider this self-recommending. (I have to confess I still like Models of Life better, but very high bar!)

2. Drugs currently in clinical trials will likely not be impacted by AI: Another good post by Abhi, this one on why drugs that are currently in clinical trials won't be (too) impacted by AI. I could quibble in some places, but agree with the overall thrusts, especially on how difficult transformation is for large, existing orgs.

3. Hundreds of Daily Decisions: Another good post by Anna Marie Wagner (co-founder of the newly announced Transfyr) and Christina Agapakis on tacit knowledge and all the little decisions people make as part of doing science.

4. Becoming a Strategist (Part 2): More Richard Rumelt, this time on becoming a strategist. I like the exercise of "analyzing current strategic situations" as both good practice and a way to escape frustration and instead think productively.

   Good strategists build such skills through practice and by a keen interest in the experiences of other leaders, companies, armies, societies, and nations. Much can be learned from other people, both privately and at gatherings. Still, there is no substitute for steadily consuming newspapers, selected articles, and books. In doing so, beware of the Internet’s memory hole: the World Wide Web first awoke in 1991, and browsers did not appear until 1993. It can be a struggle to research events that occurred before 1995.

   Once people can read books, they can educate themselves on almost anything. When I went to college in 1960, I had to read a book a week in my history and literature courses. By 2012, when I asked graduate students at UCLA to read a book, they went to the dean to complain. If schools won’t educate you, then you must educate yourself.

   Practice by Analyzing Current Strategic Situations

   Would-be strategists usually spend too much time studying strategy concepts and theories. The best way to develop the intellectual skills needed for strategy work is to practice. Short of running a company or being an advisor, the best practice is to work on a complex current situation at least twice a month.

   • Choose a current issue and identify the key forces at work—economic, social, political, and/or technological.

   • What key paradox or challenge must be resolved to move ahead?

   • What are your forecasts for how the situation will evolve over the next few months and years?

   • What would be your advice to one of the key actors?

5. How Does Claude 4 Think? — Sholto Douglas & Trenton Bricken: Sholto & Trenton back on Dwarkesh. Lots of good intuitions and near predictions about what's coming in AI. The “practitioner shop talk” genre is still extremely underrated, especially outside AI.

6. Xi Jinping’s paranoid approach to AGI, debt crisis, & Politburo politics — Victor Shih: Or, how China’s political system actually works

7. Why I don’t think AGI is right around the corner: Dwarkesh making good points about how continual learning seems underrated in the debate about transformative AI. Whether continual learning can be "bolted on" through e.g. much longer context (see the other link about math in this post), simple dynamic evaluation, and/or nonobvious but extremely simple modifications to the training procedure is one of my key uncertainties for when AI becomes really transformative. Also, credit to Steven Byrnes for tooting the horn on the importance of "figuring things out" for years now.

   I like to think I’m “AI forward” here at the Dwarkesh Podcast. I’ve probably spent over a hundred hours trying to build little LLM tools for my post production setup. And the experience of trying to get them to be useful has extended my timelines. I’ll try to get the LLMs to rewrite autogenerated transcripts for readability the way a human would. Or I’ll try to get them to identify clips from the transcript to tweet out. Sometimes I’ll try to get them to co-write an essay with me, passage by passage. These are simple, self contained, short horizon, language in-language out tasks - the kinds of assignments that should be dead center in the LLMs’ repertoire. And they're 5/10 at them. Don’t get me wrong, that’s impressive.

   But the fundamental problem is that LLMs don’t get better over time the way a human would. The lack of continual learning is a huge huge problem. The LLM baseline at many tasks might be higher than an average human's. But there’s no way to give a model high level feedback. You’re stuck with the abilities you get out of the box. You can keep messing around with the system prompt. In practice this just doesn’t produce anything even close to the kind of learning and improvement that human employees experience.

   The reason humans are so useful is not mainly their raw intelligence. It’s their ability to build up context, interrogate their own failures, and pick up small improvements and efficiencies as they practice a task.

8. Inside the Secret Meeting Where Mathematicians Struggled to Outsmart AI: Interesting anecdotes on how impressive o4-mini was when posed with novel questions (being constructed in realtime) from leading mathematicians. Not totally shocking if you've been paying attention but the anecdote here obviates a lot of the usual criticisms/caveats. Also relevant to the ongoing discussion about how hard continual learning will or won't be.

   By the end of that Saturday night, Ono was frustrated with the bot, whose unexpected mathematical prowess was foiling the group’s progress. “I came up with a problem which experts in my field would recognize as an open question in number theory—a good Ph.D.-level problem,” he says. He asked o4-mini to solve the question. Over the next 10 minutes, Ono watched in stunned silence as the bot unfurled a solution in real time, showing its reasoning process along the way. The bot spent the first two minutes finding and mastering the related literature in the field. Then it wrote on the screen that it wanted to try solving a simpler “toy” version of the question first in order to learn. A few minutes later, it wrote that it was finally prepared to solve the more difficult problem. Five minutes after that, o4-mini presented a correct but sassy solution. “It was starting to get really cheeky,” says Ono, who is also a freelance mathematical consultant for Epoch AI. “And at the end, it says, ‘No citation necessary because the mystery number was computed by me!’”
   Defeated, Ono jumped onto Signal early that Sunday morning and alerted the rest of the participants. “I was not prepared to be contending with an LLM like this,” he says, “I’ve never seen that kind of reasoning before in models. That’s what a scientist does. That’s frightening.”
   Although the group did eventually succeed in finding 10 questions that stymied the bot, the researchers were astonished by how far AI had progressed in the span of one year. Ono likened it to working with a “strong collaborator.” Yang Hui He, a mathematician at the London Institute for Mathematical Sciences and an early pioneer of using AI in math, says, “This is what a very, very good graduate student would be doing—in fact, more.”

9. Reinforcement learning and general intelligence: Finbarr Timbers, now at AI2 (big win for them), with a short but good piece on why RL is both so attractive but challenging as a path to achieving superhuman intelligence. Finbarr has put in his hours laboring in the MCTS mines so if anyone has earned the right to say RL is hard, it’s him.

10. Revenge of the junior developer: Steve Yegge on viiiiibe coding and what that means for developer jobs. Steve Yegge's blog posts, such as his platforms rant and programming "conservatives and liberals" was a pretty major inspiration for me as an early programmer who wanted to both master his craft and write well. So it's been cool to see him still writing with his same style and coming to (what I consider) very reasonable conclusions about the trajectory of AI for coding. Interesting to consider the gap between Yegge vs. AI on coding vs. writing (see also).

    All of you skeptics should drop whatever you’re doing or holding, just throw it on the ground, and run like mad towards the nearest camel and hop on. Download and try out a coding agent, ideally one launched after March 1st 2025. Because they turn whatever you know or thought you knew about coding with AI, right on its head. I myself could scarcely believe what I was seeing, just three weeks ago.

    Coding agents are simple enough in principle. They work just like a typical vibe-coding chat session, with the LLM doing most of the analysis and heavy lifting, and you mostly wearing headphones. But with agents, you don’t have to do all the ugly bidirectional copy/paste and associated prompting, which is the slow human-y part. Instead, the agent takes over and handles that for you, only returning to chat with you when it finishes or gets stuck or you run out of cash.

    And they often get pretty darn far, entirely unassisted. They just grind away at their task until they get it right, throwing tokens at the problem to explore the space as needed. The human is removed as the bottleneck for 90-99% of the work, but otherwise it’s pretty much just like a faster version of chat vibe coding.

    Every iteration of the tools from here on out will help make coding agents easier, more parallelizable, and more powerful. And we’ll start seeing truly dramatic steps forward even more often this year.

    I believe the AI-refusers regrettably have a lot invested in the status quo, which they think, with grievous mistakenness, equates to job security. They all tell themselves that the AI has yet to prove that it’s better than they are at performing X, Y, or Z, and therefore, it’s not ready yet.

    But from where I’m sitting, they’re the ones who aren’t ready. I lay this all out in detail, my friends, so you can help yourselves.

    Regardless of why the luddites aren’t adopting it, they have lost. Junior devs have the high ground, and the battle is now over. Not only are junior devs on average adopting AI faster, but junior devs are also – surprise! – cheaper. If companies are going to make cuts to pay for their devs to win with tokens, which devs do you think they’re gonna keep?

11. Software engineers are eating the world: I’ve really been enjoying David Chapman’s writings on the concept of “nobility” and this post is no exception.

    Software companies control the flow of meaning through online social networks. That gives them enormous power to shape the future of our culture, society, and selves. They also control much of the flow of goods (Amazon) and services (Facebook and Google, via advertising and influencing), giving them great economic power.

    Increasingly over the past few years, the tech industry has deliberately exercised these powers toward preferred political, cultural, psychological, and economic ends. The “tech right” claims credit for getting Trump elected, and gaining Musk unprecedented authority within the federal government.

    So. I agree that we software engineers are the wrong people for the job; but when no one else is able to do it,

    As rigid modern institutions break down, meta-rationality increasingly enables effective action. (Power!) It’s the essential postmodern capability: the creation and transformation of adaptive systems serving diverse purposes.

    The two disciplines most often engaged in practical meta-rationality are management and software engineering. In both cases, opportunities to exercise meta-rationality become available mainly only at the highest levels.

12. Superoptimization: Organizational Management By Agents: Post by cofounder of the General Intelligence Company about what manager agents may look like and how they may emerge. The email agent they discussed is one that I find particularly interesting as I keep expecting something like it to exist but haven't found something that fits what I really want.

    There are many scheduling assistants, but another manager agent that’s emerging this year is a coordinator. Using Cofounder to drive the gmail API, a memory system on your business, calendar, and slack, we’ve built an early version of a multi-person coordinator to schedule meetings and drive communication across our team.

    Here’s how it works:
    

    • Each team member has an agent with access to their email and their calendar

    • Someone requests a meeting

    • The agent reads relevant context from slack and business memory to write an agenda and determine who should be on the meeting

    • Their agent messages the other team members via email to check availability

    • The first agent sends calendar invites to relevant team members

    • The agents are optimized on calendar invites/declines

13. The Case for Bridge Editors: Niko explains better than anyone else — yes that includes the authors, but cut them some slack, they invented the damn things — has why bridge editors are a big deal.

14. The Trickiness of University Strategy: I'd been thinking about, and even prompted a few friends, with the question of, "if you put aside the frustration, what is the right strategy for universities in the face of government ire and other challenges they face?" Here, Roger Martin shares some thoughts on that Q.

    His punch line: universities should not be questioned or challenged and when they are, they should resist.

    The premise landed with a thud for me because it is not at all clear that any institution that has completely ignored criticism from outside has ever survived. “Let them eat cake” didn’t work out so hot for Marie Antoinette and French royalty in general. Politicians don’t survive by attempting to be ‘a place apart.’ Neither do political parties to which they belong. Businesses die fast when they ignore customers and/or shareholders.

    For me, the most interesting aspect of platform businesses is that the beneficiaries of the product/service the platform produces are either substantially or entirely different than the funders of the platform. Google is an example of the latter in that searchers are the beneficiaries of the product that Google provides to them — ability to search the Internet in a useful way — and they pay zero of the costs that Google incurs to make that service available. Advertisers, who are completely different entities than the searchers, pay 100% of the costs (and a lot more given the profitability of Google). In newspapers, there was (and I use ‘was’ because they no longer exist as commercial entities) a partial overlap. Readers, who are the beneficiaries of the production process of producing news, supplied about 20% of the revenues while advertisers contributed the remaining 80%.

Memorial Day Reading

When you go home, tell them of us and say,
For your tomorrow, we gave our today.
— John Maxwell Edmonds

source

On Memorial Day, I spent a couple of hours reading relevant famous speeches and essays and I found it to be quite meaningful. While Memorial Day is over, perhaps you’ll read some and get something out of them too.

1. In Flanders Fields:

   In Flanders fields the poppies blow
   Between the crosses, row on row,
   That mark our place; and in the sky
   The larks, still bravely singing, fly Scarce…

   Take up our quarrel with the foe:
   To you from failing hands we throw
   The torch; be yours to hold it high.
   If ye break faith with us who die
   We shall not sleep, though poppies grow
   In Flanders fields.

2. Oliver Wendell Holmes: "In our youths, our hearts were touched with fire", Memorial Day speech - 1884: Oliver Wendell Holmes on Memorial day.

   If this be so, the use of this day is obvious. It is true that I cannot argue a man into a desire. If he says to me, Why should I seek to know the secrets of philosophy? Why seek to decipher the hidden laws of creation that are graven upon the tablets of the rocks, or to unravel the history of civilization that is woven in the tissue of our jurisprudence, or to do any great work, either of speculation or of practical affairs? I cannot answer him; or at least my answer is as little worth making for any effect it will have upon his wishes if he asked why I should eat this, or drink that. You must begin by wanting to. But although desire cannot be imparted by argument, it can be by contagion. Feeling begets feeling, and great feeling begets great feeling. We can hardly share the emotions that make this day to us the most sacred day of the year, and embody them in ceremonial pomp, without in some degree imparting them to those who come after us. I believe from the bottom of my heart that our memorial halls and statues and tablets, the tattered flags of our regiments gathered in the Statehouses, are worth more to our young men by way of chastening and inspiration than the monuments of another hundred years of peaceful life could be.

   But as surely as this day comes round we are in the presence of the dead. For one hour, twice a year at least--at the regimental dinner, where the ghosts sit at table more numerous than the living, and on this day when we decorate their graves--the dead come back and live with us.
   I see them now, more than I can number, as once I saw them on this earth. They are the same bright figures, or their counterparts, that come also before your eyes; and when I speak of those who were my brothers, the same words describe yours.

   But, nevertheless, the generation that carried on the war has been set apart by its experience. Through our great good fortune, in our youth our hearts were touched with fire. It was given to us to learn at the outset that life is a profound and passionate thing. While we are permitted to scorn nothing but indifference, and do not pretend to undervalue the worldly rewards of ambition, we have seen with our own eyes, beyond and above the gold fields, the snowy heights of honor, and it is for us to bear the report to those who come after us. But, above all, we have learned that whether a man accepts from Fortune her spade, and will look downward and dig, or from Aspiration her axe and cord, and will scale the ice, the one and only success which it is his to command is to bring to his work a mighty heart.

   Such hearts--ah me, how many!--were stilled twenty years ago; and to us who remain behind is left this day of memories. Every year--in the full tide of spring, at the height of the symphony of flowers and love and life--there comes a pause, and through the silence we hear the lonely pipe of death. Year after year lovers wandering under the apple trees and through the clover and deep grass are surprised with sudden tears as they see black veiled figures stealing through the morning to a soldier's grave. Year after year the comrades of the dead follow, with public honor, procession and commemorative flags and funeral march--honor and grief from us who stand almost alone, and have seen the best and noblest of our generation pass away.

3. Decoration Day 88 by Robert Ingersoll:

   Nations can win success, can be rich and powerful, can cover the earth with their armies, the seas with their fleets, and yet be selfish, small and mean. Physical progress means opportunity for doing good. It means responsibility. Wealth is the end of the despicable, victory the purpose of brutality.

   But there is something nobler than all these — something that rises above wealth and power — something above lands and palaces — something above raiment and gold — it is the love of right, the cultivation of the moral nature, the desire to do justice, the inextinguishable love of human liberty.

   Nothing can be nobler than a nation governed by conscience, nothing more infamous than power without pity, wealth without honor and without the sense of justice.

   The past rises before us, and we see four millions of human beings governed by the lash — we see them bound hand and foot — we hear the strokes of cruel whips — we see the hounds tracking women through tangled swamps. We see babes sold from the breasts of mothers. Cruelty unspeakable! Outrage infinite!

   Four million bodies in chains — four million souls in fetters. All the sacred relations of wife, mother father and child trampled beneath the brutal feet of might. And all this was done under our own beautiful banner of the free.

   The past rises before us. We hear the roar and shriek of the bursting shell. The broken fetters fall. These heroes died. We look. Instead of slaves we see men and women and children. The wand of progress touches the auction block, the slave pen, the whipping post, and we see homes and firesides and school-houses and books, and where all was want and crime and cruelty and fear, we see the faces of the free.

   These heroes are dead. They died for liberty — they died for us. They are at rest. They sleep in the land they made free, under the flag they rendered stainless, under the solemn pines, the sad hemlocks, the tearful willows, and the embracing vines. They sleep beneath the shadows of the clouds, careless alike of sunshine or of storm, each in the windowless Palace of Rest. Earth may run red with other wars — they are at peace. In the midst of battle, in the roar of conflict, they found the serenity of death. I have one sentiment for soldiers living and dead: Cheers for the living; tears for the dead.

4. The Gettysburg Address by Abraham Lincoln.

5. General Douglas MacArthur: Thayer Award Address (Duty, Honor, Country).

   And 20 years after, on the other side of the globe, again the filth of murky foxholes, the stench of ghostly trenches, the slime of dripping dugouts; those boiling suns of relentless heat, those torrential rains of devastating storms; the loneliness and utter desolation of jungle trails; the bitterness of long separation from those they loved and cherished; the deadly pestilence of tropical disease; the horror of stricken areas of war; their resolute and determined defense, their swift and sure attack, their indomitable purpose, their complete and decisive victory -- always victory. Always through the bloody haze of their last reverberating shot, the vision of gaunt, ghastly men reverently following your password of: Duty, Honor, Country.

6. 1982 Memorial Day Address at Arlington National Cemetery by Ronald Reagan:

   The willingness of some to give their lives so that others might live never fails to evoke in us a sense of wonder and mystery. One gets that feeling here on this hallowed ground, and I have known that same poignant feeling as I looked out across the rows of white crosses and Stars of David in Europe, in the Philippines, and the military cemeteries here in our own land. Each one marks the resting place of an American hero and, in my lifetime, the heroes of World War I, the Doughboys, the GI's of World War II or Korea or Vietnam. They span several generations of young Americans, all different and yet all alike, like the markers above their resting places, all alike in a truly meaningful way.

   Winston Churchill said of those he knew in World War II they seemed to be the only young men who could laugh and fight at the same time. A great general in that war called them our secret weapon, "just the best darn kids in the world." Each died for a cause he considered more important than his own life. Well, they didn't volunteer to die; they volunteered to defend values for which men have always been willing to die if need be, the values which make up what we call civilization. And how they must have wished, in all the ugliness that war brings, that no other generation of young men to follow would have to undergo that same experience.

7. The Eternal Remembrance of the Brave by Pericles:

   For the whole earth is the tomb of famous men; not only are they commemorated by columns and inscriptions in their own country, but in foreign lands there dwells also an unwritten memorial of them, graven not on stone but in the hearts of men.