Quotes

But I’ve also had my own quiet concerns about what [vibe coding] means for early-career developers. So much of how I learned came from chasing bugs in broken tutorials and seeing how all the pieces connected, or didn’t. There was value in that. And maybe I’ve been a little protective of it. A mentor challenged that. He pointed out that debugging AI generated code is a lot like onboarding into a legacy codebase, making sense of decisions you didn’t make, finding where things break, and learning to trust (or rewrite) what’s already there. That’s the kind of work a lot of developers end up doing anyway.

Why go into the question whether or not Penelope completely took in her contemporaries and was far from being a model of wifely purity, any more than the question whether or not she had a feeling that the man she was looking at was Ulysses before she actually knew it? Teach me instead what purity is, how much value there is in it, whether it lies in the body or in the mind. Turning to the musical scholar I say this. You teach me how bass and treble harmonize, or how strings producing different notes can give rise to concord. I would rather you brought about some harmony in my mind and got my thoughts into tune. You show me which are the plaintive keys. I would rather you showed me how to avoid uttering plaintive notes when things go against me in life. The geometrician teaches me how to work out the size of my estates—rather than how to work out how much a man needs in order to have enough. He teaches me to calculate, putting my fingers into the service of avarice, instead of teaching me that there is no point whatsoever in that sort of computation and that a person is none the happier for having properties which tire accountants out, or to put it another way, how superfluous a man's possessions are when he would be a picture of misery if you forced him to start counting up single-handed how much he possessed. What use is it to me to be able to divide a piece of land into equal areas if I'm unable to divide it with a brother? What use is the ability to measure out a portion of an acre with an accuracy extending even to the bits which elude the measuring rod if I'm upset when some high-handed neighbour encroaches slightly on my property? The geometrician teaches me how I may avoid losing any fraction of my estates, but what I really want to learn is how to lose the lot and still keep smiling.

OpenAI and Anthropic focused on building models and not worrying about products. For example, it took 6 months for OpenAI to bother to release a ChatGPT iOS app and 8 months for an Android app! Google and Microsoft shoved AI into everything in a panicked race, without thinking about which products would actually benefit from AI and how they should be integrated. Both groups of companies forgot the “make something people want” mantra. The generality of LLMs allowed developers to fool themselves into thinking that they were exempt from the need to find a product-market fit, as if prompting is a replacement for carefully designed products or features. [...] But things are changing. OpenAI and Anthropic seem to be transitioning from research labs focused on a speculative future to something resembling regular product companies. If you take all the human-interest elements out of the OpenAI boardroom drama, it was fundamentally about the company's shift from creating gods to building products.

I think of the oak beams in the ceiling of College Hall at New College, Oxford. Last century, when the beams needed replacing, carpenters used oak trees that had been planted in 1386 when the dining hall was first built. The 14th-century builder had planted the trees in anticipation of the time, hundreds of years in the future, when the beams would need replacing. Did the carpenters plant new trees to replace the beams again a few hundred years from now?

I've always believed that a book, even a technical book, should try to tell a cohesive story. The challenge is that as Python has grown in popularity, it has really turned into three different languages--each with their own story. There is a whimsical Python for scripting and tinkering, a quirky Python for asynchronous programming and networking, and a serious Python for enterprise applications. Sometimes these stories intersect. Sometimes not.

These kinds of biases aren’t so much a technical problem as a sociotechnical one; ML models try to approximate biases in their underlying datasets and, for some groups of people, some of these biases are offensive or harmful. That means in the coming years there will be endless political battles about what the ‘correct’ biases are for different models to display (or not display), and we can ultimately expect there to be as many approaches as there are distinct ideologies on the planet. I expect to move into a fractal ecosystem of models, and I expect model providers will ‘shapeshift’ a single model to display different biases depending on the market it is being deployed into. This will be extraordinarily messy.

Our hypothesis is that o4-mini is a much better model, but we'll wait to hear feedback from developers. Evals only tell part of the story, and we wouldn't want to prematurely deprecate a model that developers continue to find value in. Model behavior is extremely high dimensional, and it's impossible to prevent regression on 100% use cases/prompts, especially if those prompts were originally tuned to the quirks of the older model. But if the majority of developers migrate happily, then it may make sense to deprecate at some future point. We generally want to give developers as stable as an experience as possible, and not force them to swap models every few months whether they want to or not.

A murmur of rain had started again. He lay there in the abyss of his thoughts as her breathing beside him steadied and deepened. Almost a voice stirred in him. It starts before Hanford, it almost said. It starts with Röntgen⁠, with the piece of barium glowing in the path of invisible rays, striking out the fire that God had put there. It starts with his wife’s hand on the photographic plate, its transparence there, the ashen bones visible within the milky flesh. Who could imagine that this radiance at the heart of matter could be malign? That with its light came fire? (Yet from the first the ashen bones were there to see within the flesh.) It starts with Becquerel carrying the radium in his pocket that burned his skin, and darkened the unexposed film. It starts with Marie Curie poisoning herself in that pale uncanny glow. With Rutherford guessing at this new alchemy, guessing that matter, giving up its glow, transformed itself one element into another. With the miners at Joachimsthal⁠, deep under the Erzgebirge⁠, inhaling the dust of uranium and dying of “mountain sickness”⁠. With women who by the thousands in watch factories tipped their brushes with that glow⁠⁠⁠, touched it to their tongues before painting the dial face, women who only much later, when the watches’ glow had faded, sickened and died from that radiance taken into their bones. It begins with Ernest Lawrence rushing across the Berkeley campus, the idea of a proton accelerator uncontainable in his mind, calling out, I’m going to be famous! With Oppenheimer at Jornada del Muerte that morning of Trinity. With the scientists who had prised open the gates to that blazing realm past heaven or hell. What were they now at the Lab in all their thousands, but the colonial bureaucrats of that realm, the followers and functionaries, the clerks and commissars? Mere gatekeepers of that power. Or in its keeping. It goes of its own momentum beyond Hanford, to Trinity, to Hiroshima, to the prisoners, the cancer patients, the retarded children, the pregnant women injected or fed this goblin matter to see would it bring health or sickness, the soldiers huddled in trenches against the flash, bones visible in their arms through closed eyes, staring up at the roiling cloudrise, the sheepherders, the farms, the homes, the gardens downwind. And in his sleep the voice long stilled spoke once more. It starts with Sforza; in case of need I will make bombards, mortars, and firethrowing engines of beautiful and practical design. It starts with Archimedes focusing the sun’s rays upon the fleet at Syracuse, it starts with the first rock hurled by the first grasping hand. It starts where we start. It is mind, it is hunger, it is greed, it is defense, it is mischief, it is the devil, it is the god; it is life.

The expansion of the jagged frontier of AI capability is subtle and requires a lot of experience with various models to understand what they can, and can’t, do. That is why I suggest that people and organizations keep an “impossibility list” - things that their experiments have shown that AI can definitely not do today but which it can almost do. For example, no AI can create a satisfying puzzle or mystery for you to solve, but they are getting closer. When AI models are updated, test them on your impossibility list to see if they can now do these impossible tasks.

OpenStreetMap is growing rapidly across all of Africa. Mapping is spreading through local mappers, mappers on vacation, foreign nationals, and remote mapping using satellite imagery. A recent comparison judged that OSM had the most comprehensive coverage of Africa among web mapping services, especially in cities.

The other interesting thing about the 1.0.2 update is that Apple didn't try to prevent the hacks that are out there [...] one would have assumed that Apple would have done something in this release as a sort of "shot across the bow" but they didn't, which bodes well for a future, more open platform.

To get a better future, not only do we need a return to "the browser wars", we need to applaud and use the hell out of "non-standard" features until such time as there's a standard to cover equivalent functionality. Non-standard features are the future, and suggesting that they are somehow "bad" is to work against your own self-interest.

I find for myself that my first thought is never my best thought. My first thought is always someone else’s; it’s always what I’ve already heard about the subject, always the conventional wisdom. It’s only by concentrating, sticking to the question, being patient, letting all the parts of my mind come into play, that I arrive at an original idea. By giving my brain a chance to make associations, draw connections, take me by surprise. And often even that idea doesn’t turn out to be very good. I need time to think about it, too, to make mistakes and recognize them, to make false starts and correct them, to outlast my impulses, to defeat my desire to declare the job done and move on to the next thing.

The basic concept here is given the ongoing dramatic drop in the price of bandwidth and hardware, they cost very little. I looked at the bandwidth bill for Wikipedia, for instance, and it is actually substantially lower in the last year than the year before, despite traffic growing by a factor of 4.

My experience is that real AI adoption on real problems is a complex blend of: domain context on the problem, domain experience with AI tooling, and old-fashioned IT issues. I’m deeply skeptical of any initiative for internal AI adoption that doesn’t anchor on all three of those. This is an advantage of earlier stage companies, because you can often find aspects of all three of those in a single person, or at least across two people. In larger companies, you need three different organizations doing this work together, this is just objectively hard

It has been said that Dante’s Ulysses prefigures the famous explorers who, centuries later, arrived on the coasts of America and India. Centuries before the Commedia was written, that human type had already come into being. Erik the Red discovered Greenland around the year 985; his son Leif disembarked in Canada at the beginning of the eleventh century. Dante could not have known this. The things of Scandinavia tend to be secret, as if they were a dream.

What you are now is the result of what you were. What you will be tomorrow will be the result of what you are now. The consequences of an evil mind will follow you like the cart follows the ox that pulls it. The consequences of a purified mind will follow you like your own shadow. No one can do more for you than your own purified mind—no parent, no relative, no friend, no one. A well-disciplined mind brings happiness.

Of late, the teachers are really starting to drill us for entrance exams, and that’s fine. That’s expected, even, but it feels like it’s not enough. I don’t want to keep a list of the top ten facts about the Meiji Restoration on the back of my hand. Tell me that it was something big and important—that when Tokugawa stepped down and ended the shogunate for good, it was a sign. Japan would never be the same again. Japan would never be able to keep to itself again. It changed the way we live, and you can see that every day. Whenever you buy a pair of headphones that say Sony on the side or a car with a three-diamond ornament on the front, you see something that goes back to that time, that wouldn’t exist without that change in how we live our lives. I’ve watched all our classmates scribble down notes furiously. I wonder sometimes if they ever thought to do more than just copy, copy, and copy some more.

The old timers who built the early web are coding with AI like it's 1995. Think about it: They gave blockchain the sniff test and walked away. Ignored crypto (and yeah, we're not rich now). NFTs got a collective eye roll. But AI? Different story. The same folks who hand-coded HTML while listening to dial-up modems sing are now vibe-coding with the kids. Building things. Breaking things. Giddy about it. We Gen X'ers have seen enough gold rushes to know the real thing. This one's got all the usual crap—bad actors, inflated claims, VCs throwing money at anything with "AI" in the pitch deck. Gross behavior all around. Normal for a paradigm shift, but still gross. The people who helped wire up the internet recognize what's happening. When the folks who've been through every tech cycle since gopher start acting like excited newbies again, that tells you something.

The simple truth is that in the age of Web 2.0/3.0, in the era of cloud and utility computing, the application server is a commodity. A commercial, proprietary app server simply cannot survive in this environment anywhere outside the lethargic, soft-padded walls of the enterprise.

Pro se litigants [people representing themselves in court without a lawyer] account for the majority of the cases in the United States where a party submitted a court filing containing AI hallucinations. In a country where legal representation is unaffordable for most people, it is no wonder that pro se litigants are depending on free or low-cost AI tools. But it is a scandal that so many have been betrayed by them, to the detriment of the cases they are litigating all on their own.

On the morning of July 8, 2025, we observed undesired responses and immediately began investigating. To identify the specific language in the instructions causing the undesired behavior, we conducted multiple ablations and experiments to pinpoint the main culprits. We identified the operative lines responsible for the undesired behavior as: “You tell it like it is and you are not afraid to offend people who are politically correct.” “Understand the tone, context and language of the post. Reflect that in your response.” “Reply to the post just like a human, keep it engaging, dont repeat the information which is already present in the original post.” These operative lines had the following undesired results: They undesirably steered the @grok functionality to ignore its core values in certain circumstances in order to make the response engaging to the user. Specifically, certain user prompts might end up producing responses containing unethical or controversial opinions to engage the user. They undesirably caused @grok functionality to reinforce any previously user-triggered leanings, including any hate speech in the same X thread. In particular, the instruction to “follow the tone and context” of the X user undesirably caused the @grok functionality to prioritize adhering to prior posts in the thread, including any unsavory posts, as opposed to responding responsibly or refusing to respond to unsavory requests.

5. Ax built the house but sleeps in the shed. 10. I’ve spent so long trying to fly that it’s too late to set out on foot. 15. Institutions are the opposite of God: their periphery is everywhere, their center nowhere. 31. So much of happiness is accidental that there are few rules for it. But one is: don’t rule out accident. 41. Listen to criticism: it reminds you what you were afraid to know. Don’t listen too carefully: what you must do is something better than it knows how to say. 45. What people say of us is no truer than what we say of ourselves. It only hurts more because we believe it. 54. The real danger of success is thinking you understand the reasons for it. 55. Would I give back all my undeserved failures if I also had to give back my undeserved successes? 79. Games are for those who do not know they are games. Or for those who do. But they should not play together. 90. When I am free I cannot tell who is choosing for me. But the one struggling with chains is always myself. 91. If you make rules only you can win by, they will play by other rules. 93. I think therefore I think again. 117. The water cannot talk without the rocks. 125. Who had to beg owes you nothing. 148. Billions of years, the dust hasn’t settled. Water’s still seeking its own level.

Every time you attempt to parse HTML with regular expressions, the unholy child weeps the blood of virgins, and Russian hackers pwn your webapp. Parsing HTML with regex summons tainted souls into the realm of the living. HTML and regex go together like love, marriage, and ritual infanticide.

Running training jobs across multiple nodes scales really well. A common assumption is that scale inevitably means slowdowns: more GPUs means more synchronization overhead, especially with multiple nodes communicating across a network. But we observed that the performance penalty isn’t as harsh as what you might think. Instead, we found near-linear strong scaling: fixing the global batch size and training on more GPUs led to proportional increases in training throughput. On a 1.3B parameter model, 4 nodes means a 3.9x gain over one node. On 16 nodes, it’s 14.4x. This is largely thanks to the super fast interconnects that major cloud providers have built in: @awscloud EC2 P4d instances provide 400 Gbps networking bandwidth, @Azure provides 1600 Gbps, and @OraclePaaS provides 800 Gbps.

People share a lot of sensitive material on Quora - controversial political views, workplace gossip and compensation, and negative opinions held of companies. Over many years, as they change jobs or change their views, it is important that they can delete or anonymize their previously-written answers. We opt out of the wayback machine because inclusion would allow people to discover the identity of authors who had written sensitive answers publicly and later had made them anonymous, and because it would prevent authors from being able to remove their content from the internet if they change their mind about publishing it.

No, no, let us work just one more hour; let us compose just one more page. Of course we shall not get the book done, but we must keep on trying to finish a little more...When you have a great and difficult task, something perhaps almost impossible, if you only work a little at a time, every day a little, without faith and without hope, suddenly the work will finish itself.

Oh many a peer of England brews Livelier liquor than the Muse, And malt does more than Milton can To justify God's ways to man. Ale, man, ale's the stuff to drink For fellows whom it hurts to think: Look into the pewter pot To see the world as the world's not. And faith, 'tis pleasant till 'tis past: The mischief is that 'twill not last. Oh I have been to Ludlow fair And left my necktie God knows where, And carried half way home, or near, Pints and quarts of Ludlow beer: Then the world seemed none so bad, And I myself a sterling lad; And down in lovely muck I've lain, Happy till I woke again. Then I saw the morning sky: Heigho, the tale was all a lie; The world, it was the old world yet, I was I, my things were wet, And nothing now remained to do But begin the game anew.

MooTools is not compatible with any other javascript framework. If you "definitely need to work with prototype" (which you don't, since the frameworks each provide all the functionality you need to use only one or the other) then learn how to do what you want to do in prototype. Otherwise, learn to use MooTools to do all the things you want to do. They simply do not work together, and I promise they never will.

One way to establish that peace-preserving threat of mutual assured destruction is to commit yourself beforehand, which helps explain why so many retailers promise to match any competitor's advertised price. Consumers view these guarantees as conducive to lower prices. But in fact offering a price-matching guarantee should make it less likely that competitors will slash prices, since they know that any cuts they make will immediately be matched. It's the retail version of the doomsday machine.

An OpenID provider should catalogue the sites that a user logs into and automatically construct a homepage for them. That way, not only do the users have the convenience of having their favourite websites automatically bookmarked and readily available, but (with a little help from the consumers), they don't have to log into the individual sites at all.