Quotes

in July 2023, we [Hugging Face] wanted to experiment with a custom license for this specific project [text-generation-inference] in order to protect our commercial solutions from companies with bigger means than we do, who would just host an exact copy of our cloud services. The experiment however wasn't successful. It did not lead to licensing-specific incremental business opportunities by itself, while it did hamper or at least complicate the community contributions, given the legal uncertainty that arises as soon as you deviate from the standard licenses.

Perhaps the excellence of aphorisms consists not so much in the expression of some rare or abstruse sentiment, as in the comprehension of some obvious and useful truth in a few words. We frequently fall into error and folly, not because the true principles of action are not known, but because, for a time, they are not remembered; and he may therefore be justly numbered among the benefactors of mankind who contracts the great rules of life into short sentences, that may be easily impressed on the memory, and taught by frequent recollection to recur habitually to the mind.

Async functions require an event loop to run. Flask, as a WSGI application, uses one worker to handle one request/response cycle. When a request comes in to an async view, Flask will start an event loop in a thread, run the view function there, then return the result. Each request still ties up one worker, even for async views. The upside is that you can run async code within a view, for example to make multiple concurrent database queries, HTTP requests to an external API, etc. However, the number of requests your application can handle at one time will remain the same.

Q: Why does the devil keep his deals? A: Because as an immortal he has an infinite time horizon of other deals he jeopardizes if he betrays any given deal. Therefore the opportunity cost of any betrayal is too high. Q: What does that make politicians then? A: Lower in ethical reliability than the devil.

At the start of the year, most people loosely following AI probably knew of 0 [Chinese] AI labs. Now, and towards wrapping up 2025, I’d say all of DeepSeek, Qwen, and Kimi are becoming household names. They all have seasons of their best releases and different strengths. The important thing is this’ll be a growing list. A growing share of cutting edge mindshare is shifting to China. I expect some of the likes of Z.ai, Meituan, or Ant Ling to potentially join this list next year. For some of these labs releasing top tier benchmark models, they literally started their foundation model effort after DeepSeek. It took many Chinese companies only 6 months to catch up to the open frontier in ballpark of performance, now the question is if they can offer something in a niche of the frontier that has real demand for users.

[...] if your situation allows it, always try uv first. Then fall back on something else if that doesn’t work out. It is the Pareto solution because it's easier than trying to figure out what you should do and you will rarely regret it. Indeed, the cost of moving to and from it is low, but the value it delivers is quite high.

I've worked out why I don't get much value out of LLMs. The hardest and most time-consuming parts of my job involve distinguishing between ideas that are correct, and ideas that are plausible-sounding but wrong. Current AI is great at the latter type of ideas, and I don't need more of those.

Most classical engineering fields deal with probabilistic system components all of the time. In fact I'd go as far as to say that inability to deal with probabilistic components is disqualifying from many engineering endeavors. Process engineers for example have to account for human error rates. On a given production line with humans in a loop, the operators will sometimes screw up. Designing systems to detect these errors (which are highly probabilistic!), mitigate them, and reduce the occurrence rates of such errors is a huge part of the job. [...] Software engineering is unlike traditional engineering disciplines in that for most of its lifetime it's had the luxury of purely deterministic expectations. This is not true in nearly every other type of engineering.

That the pleasure arising to man from contact with sensible objects, is to be relinquished because accompanied by pain--- such is the reasoning of fools. The kernels of the paddy, rich with finest white grains, What man, seeking his own true interest, would fling them away because of a covering of husk and dust?

Government in the UK once lead the world in it's own information systems, breaking Enigma, documenting an empire's worth of trade. And then it fired everyone who could do those things, or employed them only via horribly expensive consultancies. It is time to start bringing them back into the corridors of power.

A new, a vast, and a powerful language is developed for the future use of analysis, in which to wield its truths so that these may become of more speedy and accurate practical application for the purposes of mankind than the means hitherto in our possession have rendered possible.

Long before nightfall someone who exchanged greetings with darkness comes to spend the night with you. Long before daylight he wakes and, before leaving, kindles a sleep, a sleep echoing with footsteps: you hear him going off, measuring distances, and you throw your soul after him.

Did I Cry, why? Many times. To no answer But the ungentle angels. Finally I understood; and some time later, Was smelted of my crimes. My soul Was good silver, thinner than paper, Stamp-small; such remained; such Big angels brought before the Lord. Who spoke formalities, and sent me Back to the one hell, this earth, Where it is summer always, where men Are animals, where I whisper Like a leaf in the perfect breeze.

There are... many types of monsters in this world: Monsters who will not show themselves and who cause trouble; monsters who abduct children; monsters who devour dreams; monsters who suck blood, and... monsters who always tell lies. Lying monsters are a real nuisance. They are much more cunning than other monsters. They pose as humans even though they have no understanding of the human heart. They eat even though they've never experienced hunger. They study even though they have no interest in academics. They seek friendship even though they do not know how to love. If I were to encounter such a monster, I would likely be eaten by it. Because in truth, I am that monster.

I started using Claude and Claude Code a bit in my regular workflow. I’ll skip the suspense and just say that the tool is way more capable than I would ever have expected. The way I can use it to interrogate a large codebase, or generate unit tests, or even “refactor every callsite to use such-and-such pattern” is utterly gobsmacking. [...] Here’s the main problem I’ve found with generative AI, and with “vibe coding” in general: it completely sucks out the joy of software development for me. [...] This is how I feel using gen-AI: like a babysitter. It spits out reams of code, I read through it and try to spot the bugs, and then we repeat.

I thought that everyone in Japan had to be packed in there. So I turned to my dad and asked him, "Do you know how many people are here right now"? He said since the stadium was full, probably fifty thousand...I was only one little person in that big crowded stadium filled with people, and there were so many people there, but it was just a handful out of the entire population. Up till then, I always thought that I was, I don’t know, kind of a special person. It was fun to be with my family. I had fun with my classmates. And the school that I was going to, it had just about the most interesting people anywhere. But that night, I realized it wasn’t true. All the stuff we did during class that I thought was so fun and cool, was probably happening just like that in classes in other schools all over Japan. There was nothing special about my school at all. ...If there’s really that many people in the world, then there had to be someone who wasn’t ordinary. There had to be someone who was living an interesting life. There just had to be. Why wasn’t I that person?

We used this model [periodically transmitting configuration to different hosts] to distribute translations, feature flags, configuration, search indexes, etc at Airbnb. But instead of SQLite we used Sparkey, a KV file format developed by Spotify. In early years there was a Cron job on every box that pulled that service’s thingies; then once we switched to Kubernetes we used a daemonset & host tagging (taints?) to pull a variety of thingies to each host and then ensure the services that use the thingies only ran on the hosts that had the thingies.

Are we so deranged here in the twenty-first century that we’re going to re-enact, wide-eyed, the twin tragedies of the great desktop-suite lock-in and the great proprietary-SQL lock-in? You know, the ones where you give a platform vendor control over your IT budget? Gimme a break.

Simplicity of life, even the barest, is not a misery, but the very foundation of refinement: a sanded floor and whitewashed walls, and the green trees, and flowery meads, and living waters outside; or a grimy palace amid the smoke with a regiment of housemaids always working to smear the dirt together so that it may be unnoticed; which, think you, is the most refined, the most fit for a gentleman of those 2 dwellings?

That development time acceleration of 4 days down to 20 minutes… that’s equivalent to about 10 years of Moore’s Law cycles. That is, using generative AI like this is equivalent to computers getting 10 years better overnight. That was a real eye-opening framing for me. AI isn’t magical, it’s not sentient, it’s not the end of the world nor our saviour; we don’t need to endlessly debate “intelligence” or “reasoning.” It’s just that… computers got 10 years better. The iPhone was first released in 2007. Imagine if it had come out in 1997 instead. We wouldn’t even know what to do with it.

When many business people talk about “AI” today, they treat it as a continuum with past capabilities of the CNN/RNN/GAN world. In reality it is a step function in new capabilities and products enabled, and marks the dawn of a new era of tech. It is almost like cars existed, and someone invented an airplane and said “an airplane is just another kind of car - but with wings” - instead of mentioning all the new use cases and impact to travel, logistics, defense, and other areas. The era of aviation would have kicked off, not the “era of even faster cars”.

In practice replacing digital computers with an alternative computing paradigm is a risky proposition. Alternative computing architectures, such as parallel digital computers have not tended to be commercially viable, because Moore’s Law has consistently enabled conventional von Neumann architectures to render alternatives unnecessary. Besides Moore’s Law, digital computing also benefits from mature tools and expertise for optimizing performance at all levels of the system: process technology, fundamental circuits, layout and algorithms. Many engineers are simultaneously working to improve every aspect of digital technology, while alternative technologies like analog computing do not have the same kind of industry juggernaut pushing them forward.

EVE Online is unique in gaming in that we have always played on the same massive server in the same online universe since May 2003 when it first went live. We not only understand the harsh penalties for failure, but also how longevity and persistence is rewarded with success. When you have over 60,000 people on weekends dealing, scheming, and shooting each other, it attracts a certain type of gamer. It’s not a quick fix kind of game. We enjoy building things that last, be they virtual spaceships or real life friendships that together translate into massive Empires and enduring legacies. Those of us who play understand that one man really can truly make a difference in our world.

By cutting out a hundred voices or fewer, things and people that everybody talks about became things and people that nobody talks about. The internet is a technology for creating small ponds for us to all be big fish in. But you change your perspective just slightly, move over just an inch, and suddenly you get a sense of just how few people know about you or could possibly care.

Each speaker gets five minutes to explain their research, with a human metronome banging a waste bin with a big stick after every minute. After five minutes, an eight-year old girl (last night, actually two twins) walks across the stage and says "Please Stop, I'm Bored" and repeats it until the speaker does indeed stop.

A McAfee spokeswoman said the company rates XSS vulnerabilities less severe than SQL injections and other types of security bugs. "Currently, the presence of an XSS vulnerability does not cause a web site to fail HackerSafe certification," she said. "When McAfee identifies XSS, it notifies its customers and educates them about XSS vulnerabilities."

The way I would talk about myself as a senior engineer is that I’d say “I know how I would solve the problem” and because I know how I would solve it I could also teach someone else to do it. And my theory is that the next level is that I can say about myself “I know how others would solve the problem”. Let’s make that a bit more concrete. You make that sentence: “I can anticipate how the API choices that I’m making, or the abstractions that I’m introducing into a project, how they impact how other people would solve a problem.”

3. "He abused me, he beat me, he defeated me, he robbed me",—in those who harbour such thoughts hatred will never cease. 4 "He abused me, he beat me, he defeated me, he robbed me",—in those who do not harbour such thoughts hatred will cease. 5 For hatred does not cease by hatred at any time: hatred ceases by love, this is an old rule.

Fear keeps pace with hope. Nor does their so moving together surprise me; both belong to a mind in suspense, to a mind in a state of anxiety through looking into the future. Both are mainly due to projecting our thoughts far ahead of us instead of adapting ourselves to the present. Thus it is that foresight, the greatest blessing humanity has been given, is transformed into a curse. Wild animals run from the dangers they actually see, and once they have escaped them worry no more. We however are tormented alike by what is past and what is to come. A number of our blessings do us harm, for memory brings back the agony of fear while foresight brings it on prematurely. No one confines his unhappiness to the present.

Once you reach a certain level of activity in the system where the garbage collector can no longer keep up (and it will happen), then every line of code in your system is now a potential failure point that can leave the whole program in a bad state. Lisp has this problem. Java has this problem. Erlang does not.