Quotes

I designed Dropbox's storage system and modeled its durability. Durability numbers (11 9's etc) are meaningless because competent providers don't lose data because of disk failures, they lose data because of bugs and operator error. [...] The best thing you can do for your own durability is to choose a competent provider and then ensure you don't accidentally delete or corrupt own data on it: 1. Ideally never mutate an object in S3, add a new version instead. 2. Never live-delete any data. Mark it for deletion and then use a lifecycle policy to clean it up after a week. This way you have time to react to a bug in your own stack.

A successful author is equally in danger of the diminution of his fame, whether he continues or ceases to write. The regard of the publick is not to be kept but by tribute, and the remembrance of past service will quickly languish, unless successive performances frequently revive it. Yet in every new attempt there is new hazard, and there are few who do not at some unlucky time, injure their own characters by attempting to enlarge them.

It seems as if you are never ‘hardcore’ enough for YouTube’s recommendation algorithm. It promotes, recommends and disseminates videos in a manner that appears to constantly up the stakes. Given its billion or so users, YouTube may be one of the most powerful radicalising instruments of the 21st century.

My apologies to chance for calling it necessity. My apologies to necessity if I’m mistaken, after all. Please, don’t be angry, happiness, that I take you as my due. May my dead be patient with the way my memories fade. My apologies to time for all the world I overlook each second. My apologies to past loves for thinking that the latest is the first. Forgive me, distant wars, for bringing flowers home. Forgive me, open wounds, for pricking my finger. I apologize for my record of minuets to those who cry from the depths. I apologize to those who wait in railway stations for being asleep today at 5AM Pardon me, hounded hope, for laughing from time to time. Pardon me, deserts, that I don’t rush to you bearing a spoonful of water. And you, falcon, unchanging year after year, always in the same cage, your gaze always fixed on the same point in space, forgive me, even if it turns out you were stuffed. My apologies to the felled tree for the table’s four legs. My apologies to great questions for small answers. ...Don’t bear me ill will, speech, that I borrow weighty words, then labor heavily so that they may seem light.

In Python 3.14, I have implemented several changes to fix thread safety of `asyncio` and enable it to scale effectively on the free-threaded build of CPython. It is now implemented using lock-free data structures and per-thread state, allowing for highly efficient task management and execution across multiple threads. In the general case of multiple event loops running in parallel, there is no lock contention and performance scales linearly with the number of threads. [...] For a deeper dive into the implementation, check out the internal docs for asyncio.

Rye lets you get from no Python on a computer to a fully functioning Python project in under a minute with linting, formatting and everything in place. [...] Because it was demonstrably designed to avoid interference with any pre-existing Python configurations, Rye allows for a smooth and gradual integration and the emotional barrier of picking it up even for people who use other tools was shown to be low.

The process of learning and experimenting with LLM-derived technology has been an exercise in humility. In general I love learning new things when the art of programming changes […] But LLMs, and more specifically Agents, affect the process of writing programs in a new and confusing way. Absolutely every fundamental assumption about how I work has to be questioned, and it ripples through all the experience I have accumulated. There are days when it feels like I would be better off if I did not know anything about programming and started from scratch. And it is still changing.

Most propositions & questions that have been written about philosophical matters are not false but senseless. We cannot, therefore, answer questions of this kind at all, but only state their senselessness. Most questions & propositions of the philosophers result from the fact that we do not understand the logic of our language. They are of the same kind as the question whether the Good is more, or less, identical than the Beautiful.

Voters in the Clapham and Brixton Hill constituency can rest easy - despite appearances, their Reform candidate Mark Matlock really does exist. [...] Matlock - based in the South Cotswolds, some 100 miles from the constituency in which he is standing - confirmed: "I am a real person." Although his campaign image is Al-generated, he said this was for lack of a real photo of him wearing a tie in Reform's trademark turquoise.

What Apple unveiled last week with Apple Intelligence wasn't so much new products, but new features—a slew of them—for existing products, powered by generative AI. [...] These aren't new apps or new products. They're the most used, most important apps Apple makes, the core apps that define the Apple platforms ecosystem, and Apple is using generative AI to make them better and more useful—without, in any way, rendering them unfamiliar.

Tuesday’s chaos arose after China Railway Shenyang failed to deactivate Flash in time, leading to a complete shutdown of its railroads in Dalian, Liaoning province. Staffers were reportedly unable to view train operation diagrams, formulate train sequencing schedules and arrange shunting plans. Authorities fixed the issue by installing a pirated version of Flash at 4:30 a.m. the following day.

Apart from which this kind of obsession with the liberal arts turns people into pedantic, irritating, tactless, self-satisfied bores, not learning what they need simply because they spend their time learning things they will never need. The scholar Didymus wrote four thousand works: I should feel sorry for him if he had merely read so many useless works. In these works he discusses such questions as Homer’s origin, who was Aeneas’ real mother, whether Anacreon’s manner of life was more that of a lecher or that of a drunkard, whether Sappho slept with anyone who asked her, and other things that would be better unlearned if one actually knew them! Don’t you go and tell me now that life is long enough for this sort of thing!

Goddammit. The Onion once again posted an article in which a portion of the artwork came from an AI-generated Shutterstock image. This article was over a month old and only a portion of the image. We took it down immediately. [...] To be clear, The Onion has a several-person art team and they work their asses off. Sometimes they work off of stock photo bases and go from there. That's what happened this time. This was not a problem until stock photo services became flooded with AI slop. We'll reinforce process and move on.

One of the things we did all the time at early GitHub was a two-step ship: basically, ship a big launch, but days or weeks afterwards, ship a smaller, add-on feature. In the second launch post, you can refer back to the initial bigger post and you get twice the bang for the buck. This is even more valuable than on the surface, too: you get to split your product launch up into a few different pieces, which lets you slowly ease into the full usage — and server load — of new code.

It is hard to communicate how much programming has changed due to AI in the last 2 months: not gradually and over time in the "progress as usual" way, but specifically this last December. There are a number of asterisks but imo coding agents basically didn’t work before December and basically work since - the models have significantly higher quality, long-term coherence and tenacity and they can power through large and long tasks, well past enough that it is extremely disruptive to the default programming workflow. [...]

Following the widespread availability of large language models (LLMs), the Django Security Team has received a growing number of security reports generated partially or entirely using such tools. Many of these contain inaccurate, misleading, or fictitious content. While AI tools can help draft or analyze reports, they must not replace human understanding and review. If you use AI tools to help prepare a report, you must: - Disclose which AI tools were used and specify what they were used for (analysis, writing the description, writing the exploit, etc). - Verify that the issue describes a real, reproducible vulnerability that otherwise meets these reporting guidelines. - Avoid fabricated code, placeholder text, or references to non-existent Django features. Reports that appear to be unverified AI output will be closed without response. Repeated low-quality submissions may result in a ban from future reporting

The essence of magic is to do away with underlying mechanisms...What makes the elephant disappear is the movement of the wand and the intent of the magician, directly. If there were any intervening processes, it would not be magic but just engineering. As soon as you know how the magician made the elephant disappear, the magic disappears and—if you started by believing in magic—the disappointment sets in.

Thou art a dreaming thing, A fever of thyself think of the Earth; What bliss even in hope is there for thee? What haven? every creature hath its home; Every sole man hath days of joy and pain, Whether his labours be sublime or low The pain alone; the joy alone; distinct: Only the dreamer venoms all his days, Bearing more woe than all his sins deserve.

—Suddenly Einstein stopped reading and handed me a telegram he had picked up from the window sill, saying, “You might be interested in this.” It was a telegram from Eddington with the results of the famous eclipse expedition. Full of excitement, I exclaimed, “How wonderful, this is almost the value you calculated.” Completely calm, he remarked, “I knew the theory was correct. Did you doubt it?” I replied, “No, of course not. But what would you have said if the confirmation hadn’t been so?” He replied: “I could just feel sorry for dear God, the theory is correct.” Here he used—as so often—the word “God” instead of “nature”.

If you work slowly, you will be more likely to stick with your slightly obsolete work. You know that professor who spent seven years preparing lecture notes twenty years ago? He is not going to throw them away and start again, as that would be a new seven-year project. So he will keep teaching using aging lecture notes until he retires and someone finally updates the course.

There are no fixtures in nature. The universe is fluid and volatile. Permanence is but a word of degrees. Our globe seen by God is a transparent law, not a mass of facts. The law dissolves the fact and holds it fluid. Our culture is the predominance of an idea which draws after it this train of cities and institutions. Let us rise into another idea: they will disappear. …The new continents are built out of the ruins of an old planet; the new races fed out of the decomposition of the foregoing. New arts destroy the old. See the investment of capital in aqueducts made useless by hydraulics; fortifications, by gunpowder; roads and canals, by railways; sails, by steam; steam, by electricity. You admire this tower of granite, weathering the hurts of so many ages. Yet a little waving hand built this huge wall, and that which builds is better than that which is built. The hand that built can topple it down much faster. Better than the hand, and nimbler, was the invisible thought which wrought through it, and thus ever behind the coarse effect is a fine cause which, being narrowly seen, is itself the effect of a finer cause. Everything looks permanent until its secret is known.

“No doubt you are right, Mr. Curtis”, said Andre, smiling, “but poets are like proverbs; you can always find one to contradict another. Although Waller and Moore have chosen to sing the praises of the Bermudas, it has been supposed that Shakespeare was depicting them in the terrible scenes that are found in The Tempest.”

[On Meta's Galactica LLM launch] We did this with a 8 person team which is an order of magnitude fewer people than other LLM teams at the time. We were overstretched and lost situational awareness at launch by releasing demo of a base model without checks. We were aware of what potential criticisms would be, but we lost sight of the obvious in the workload we were under. One of the considerations for a demo was we wanted to understand the distribution of scientific queries that people would use for LLMs (useful for instruction tuning and RLHF). Obviously this was a free goal we gave to journalists who instead queried it outside its domain. But yes we should have known better. We had a “good faith” assumption that we’d share the base model, warts and all, with four disclaimers about hallucinations on the demo - so people could see what it could do (openness). Again, obviously this didn’t work.

Often, you are told to do this by treating AI like an intern. In retrospect, however, I think that this particular analogy ends up making people use AI in very constrained ways. To put it bluntly, any recent frontier model (by which I mean Claude 3.5, ChatGPT-4o, Grok 2, Llama 3.1, or Gemini Pro 1.5) is likely much better than any intern you would hire, but also weirder. Instead, let me propose a new analogy: treat AI like an infinitely patient new coworker who forgets everything you tell them each new conversation, one that comes highly recommended but whose actual abilities are not that clear.

You pass briefly through our midst. Your books and writing have not been consulted. Our prayers are pro defuncto N. Yet some look up, as though among a crowd of prisoners or displaced persons, they recognized a friend once known in a far country. For these the sun also rose after a forgotten war upon an idiom you made great. They have not forgotten you. In their silence you are still famous, no ritual shade. How slowly this bell tolls in a monastery tower for a whole age, and for the quick death of an unready dynasty, and for that brave illusion: the adventurous self? For with one shot the whole hunt is ended!

This is what I constantly tell my students: The hard part about doing a tech product for the most part isn't the what beginners think makes tech hard — the hard part is wrangling systemic complexity in a good, sustainable and reliable way. Many non-tech people e.g. look at programmers and think the hard part is knowing what this garble of weird text means. But this is the easy part. And if you are a person who would think it is hard, you probably don't know about all the demons out there that will come to haunt you if you don't build a foundation that helps you actively keeping them away.

I think most people have this naive idea of consensus meaning “everyone agrees”. That’s not what consensus means, as practiced by organizations that truly have a mature and well developed consensus driven process. Consensus is not “everyone agrees”, but [a model where] people are more aligned with the process than they are with any particular outcome, and they’ve all agreed on how decisions will be made.

OMNI: Will robots be complex enough to be friends of people, do you think? Shannon: I think so. But it’s quite a way away. OMNI: Could you imagine being friends with a robot? Shannon: Yes I could. I could imagine that happening very easily. I see no limit to the capability of machines. The microchips are getting smaller and smaller and faster and faster and I can see them getting better than we are. I can visualize sometime in the future we will be to robots as dogs are to humans.