Dan Luu
Programmer and writer on computer architecture, performance, and software reliability. He has worked on CPU design at Centaur Technology and on software at Google and Microsoft, and writes long-form technical essays at danluu.com.
Dan Luu did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
Latest x.com
What they believe
-
Even at top-100 traffic levels most applications can be served by a simple monolith, and it is cheaper to build than a complex architecture.
computers are fast enough that high-traffic apps can be served with simple architectures, which can generally be created more cheaply and easily than complex architectures
danluu.comSaid 6 Apr 2022This on korrents
-
People reliably skip the practices known to prevent rare catastrophes, and then rationalise the shortcut as the reasonable thing to do.
The data are clear that humans are really bad at taking the time to do things that are well understood to incontrovertibly reduce the risk of rare but catastrophic events.
danluu.comSaid 29 Dec 2015This on korrents
The bold line is korrents’ wording; the quote is theirs. These 2: what they wrote before what they said aloud, one per source, newest first.
Feed
As its own page →Hiding
16 October
2 October
28 September
25 September
23 September
22 September
18 September
16 September
15 September
10 September
7 September
4 September
1 September
29 August
24 August
21 August
17 August
10 August
Nothing matches.
Beliefs
Korrents What they believe 15 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
Even at top-100 traffic levels most applications can be served by a simple monolith, and it is cheaper to build than a complex architecture.
computers are fast enough that high-traffic apps can be served with simple architectures, which can generally be created more cheaply and easily than complex architectures
In defense of simple architectures Said 6 Apr 2022
Conference programmes and the tech press over-represent complex architectures, so teams copy techniques their scale does not need.
Despite the unreasonable effectiveness of simple architectures, most press goes to complex architectures.
In defense of simple architectures Said 6 Apr 2022
The cost of the engineering team dominates the cost of the machines it runs, so a slow language is usually an affordable choice.
The cost of our engineering team completely dominates the cost of the systems we operate
In defense of simple architectures Said 6 Apr 2022
Show 12 more
People reliably skip the practices known to prevent rare catastrophes, and then rationalise the shortcut as the reasonable thing to do.
The data are clear that humans are really bad at taking the time to do things that are well understood to incontrovertibly reduce the risk of rare but catastrophic events.
Normalization of deviance Said 29 Dec 2015
Normalization of deviance is well studied in healthcare and aviation and almost never discussed in software, although software organisations do it constantly.
It's well studied in a number of other contexts including healthcare, aviation, mechanical engineering, aerospace engineering, and civil engineering, but we don't see it discussed in the context of software.
Normalization of deviance Said 29 Dec 2015
Writing to files safely is too hard to do with raw syscalls; application code should go through a library or a database.
as an application developer, writing to files safely is hard enough that it should be done via some kind of library and/or database, not by directly making syscalls
Files are hard Said 12 Dec 2015
Almost every software project has cheap testing wins left on the table: basic random testing, static analysis and fault injection pay for themselves the first time they are run.
Almost every software project I've seen has a lot of low hanging testing fruit.
Files are hard Said 12 Dec 2015
Conventional testing is not sufficient: crude purpose-built checking tools of a few thousand lines find bugs that testing does not.
These studies also hammer home the point that conventional testing isn't sufficient.
Files are hard Said 12 Dec 2015
A working programmer routinely works around hundreds of bugs a week in ordinary software, and the industry has normalised it.
If I had to guess, I'd say I probably work around hundreds of bugs in an average week, and thousands in a bad week.
Everything is broken Said 18 Nov 2014
How much to test is an economic calculation a team should actually run, not a matter of principle.
The point isn't that you should definitely write more tests, it's that you should definitely do the math to see if you should write more tests.
Everything is broken Said 18 Nov 2014
The empirical literature on static typing finds effects too small to matter, in the narrow settings it has managed to study.
The summary of the summary is that most studies find very small effects, if any.
Literature review on the benefits of static types Said 7 Nov 2014
He prefers languages with expressive type systems and has still never noticed them making him more productive or less bug-prone.
But while I like languages with expressive type systems, I haven't noticed that they make me more productive or less bug prone
Literature review on the benefits of static types Said 7 Nov 2014
"There is no empirical evidence against my position" is not an argument for it.
Maybe, but if the strongest statement you can make for your position is that there's no empirical evidence against the position, that's not much of a position.
Literature review on the benefits of static types Said 7 Nov 2014
The answer to weak studies of programming practice is to fund better ones, not to conclude that the question cannot be measured.
It's true that controlled studies only tell you something about a very limited set of circumstances, but the fix to that isn't to dismiss them, but to fund more studies.
Literature review on the benefits of static types Said 7 Nov 2014
The most-cited study showing that language design affects software quality rests on misclassified repositories.
So the intermediate result appears to not be that TypeScript is reliable, but that projects mis-identified as TypeScript are reliable.
Literature review on the benefits of static types Said 7 Nov 2014
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Recommends rcmnd
rcmnd What they recommend 31 things — each backed by an exact quote.
Compiled by rcmnd.app, not by them: the quotes are theirs, and each links to the post it was copied from.
Recent
Liked Renegades of the Empire
This book explains how Microsoft's aggressive culture got to be the way it is today.
danluu.com 1 Oct 2016
affiliate link
Liked Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations
The title is the worst thing about this book. Otherwise, it's a nice introduction to algorithmic game theory.
danluu.com 1 Oct 2016
affiliate link
What others say about Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations →
Liked Snipers, Shills, and Sharks: eBay and Human Behavior
Seems designed as an entertaining introduction to auction theory for the layperson.
danluu.com 1 Oct 2016
affiliate link
What others say about Snipers, Shills, and Sharks: eBay and Human Behavior →
Show 28 more
Loved Computation: Finite and Infinite Machines
The explanations are delightful and I've watched some of the videos more than once just to watch Bernhardt explain things.
danluu.com 1 Oct 2016
affiliate link
What others say about Computation: Finite and Infinite Machines →
Loved Probability and Computing: Randomized Algorithms and Probabilistic Analysis
I've probably gotten more mileage out of this than out of any other algorithms book.
danluu.com 1 Oct 2016
affiliate link
What others say about Probability and Computing: Randomized Algorithms and Probabilistic Analysis →
Mixed on Peopleware
This book agrees with my biases and I'd love for this book to be right, but the meta evidence makes me want to re-read this with a critical eye and look up primary sources.
danluu.com 1 Oct 2016
affiliate link
Liked Refactoring
At the time I read it, it was worth the price of admission for the section on code smells alone.
danluu.com 1 Oct 2016
affiliate link
Mixed on Site Reliability Engineering
Liked Multiagent Systems
The title is the worst thing about this book. Otherwise, it's a nice introduction to algorithmic game theory.
danluu.com 1 Oct 2016
Liked Combinatorial Auctions
This is one of those books where each chapter is by a different author. Despite that, it still manages to be coherent and I didn't mind reading it straight through.
danluu.com 1 Oct 2016
Liked Snipers, Shills, and Sharks
Seems designed as an entertaining introduction to auction theory for the layperson. Requires no mathematical background and relegates math to the small print.
danluu.com 1 Oct 2016
affiliate link
Liked Auction Theory
The last time I looked, this was the only game in town for a comprehensive, modern, introduction to auction theory.
danluu.com 1 Oct 2016
affiliate link
Loved Modern Processor Design
For pure CPU architecture, this is probably the best book available.
danluu.com 1 Oct 2016
affiliate link
Liked Computer Architecture: A Quantitative Approach
This book teaches you how to do systems design with multiple constraints (e.g., performance, TCO, and power) and how to reason about tradeoffs.
danluu.com 1 Oct 2016
What others say about Computer Architecture: A Quantitative Approach →
Liked The Little Book of Semaphores
This book was very helpful when I first started to write threading/concurrency code.
danluu.com 1 Oct 2016
Liked Windows Internals
The most comprehensive book about how a modern operating system works. It just happens to be about Windows.
danluu.com 1 Oct 2016
affiliate link
Mixed on Linux Kernel Development
The other downside of this book is that the author loses all objectivity any time Linux and Windows are compared.
danluu.com 1 Oct 2016
affiliate link
Liked Operating Systems: Three Easy Pieces
Nice explanation of a variety of OS topics. Goes into much more detail than any other intro OS book I know of.
danluu.com 1 Oct 2016
affiliate link
What others say about Operating Systems: Three Easy Pieces →
Loved xv6
This book is great! It explains how you can actually implement things in a real system
danluu.com 1 Oct 2016
Mixed on Operating System Concepts
It covers concepts at a high level and hits the major points, but it's lacking in technical depth, details on how things work, advanced topics, and clear exposition.
danluu.com 1 Oct 2016
affiliate link
Disliked An Introduction to Computational Learning Theory
Classic, but dated and riddled with errors, with no errata available.
danluu.com 1 Oct 2016
affiliate link
What others say about An Introduction to Computational Learning Theory →
Loved Computation
Covers a few theory of computation highlights. The explanations are delightful and I've watched some of the videos more than once just to watch Bernhardt explain things.
danluu.com 1 Oct 2016
Mixed on Introduction to the Theory of Computation
Unfortunately, most of the key exercises don't have solutions, so you can't check your work.
danluu.com 1 Oct 2016
affiliate link
What others say about Introduction to the Theory of Computation →
Loved Probability and Computing
I've probably gotten more mileage out of this than out of any other algorithms book.
danluu.com 1 Oct 2016
affiliate link
Recommends Higher Order Perl
A gentle introduction to functional programming that happens to use Perl. You could probably work through this book just as easily in Python or Ruby.
danluu.com 1 Oct 2016
Mixed on Purely Functional Data Structures
Fun to work through, but, unlike the other algorithms and data structures books, I've yet to be able to apply anything from this book to a problem domain where performance really matters.
danluu.com 1 Oct 2016
affiliate link
Mixed on Algorithm Design
Same comments as for CLRS -- it's widely recommended as an introductory book even though it doesn't make sense as an introductory book.
danluu.com 1 Oct 2016
affiliate link
Mixed on Introduction to Algorithms
but this is wildly impractical as a standalone introduction for most people.
danluu.com 1 Oct 2016
affiliate link
Liked The Algorithm Design Manual
The longer, more comprehensive, more practical, less math-y version of Dasgupta. It's similar in that it attempts to teach you how to identify problems, use the correct algorithm, and give a clear explanation of the algorithm.
danluu.com 1 Oct 2016
affiliate link
Loved Algorithms
Everything about this book seems perfect to me. It breaks up algorithms into classes (e.g., divide and conquer or greedy), and teaches you how to recognize what kind of algorithm should be used to solve a particular problem.
danluu.com 1 Oct 2016
affiliate link
Mixed on Cracking the Coding Interview
I usually recommend this book to people who want to pass interviews but not really learn about algorithms. It has just enough to get by, but doesn't really teach you the why behind anything.
danluu.com 1 Oct 2016
affiliate link
Where they publish
x.com @danluu His older posting; the account is linked from every page of his site.
Recent
- Friend of mine started at a lab recently. In orientation, they warned him that, if his affiliation was public, fake identities would reach out to him to extract information. This happened multiple times in his first week, exactly as described in the orientation warning. 16 Oct 2026
- At the time, I did automated bug filing with human review. Human review found zero known false positives. Anthropic was using a better model and had more tokens to spend, giving them the advantage on the two most difficult factors w.r.t. false positive rejection. 23 Sept 2026
- Why have Anthropic's Mythos/Glasswing vuln reports been so bad? In https://t.co/JmS2TEug3f, I mentioned a colleague saying the reports we got were mostly nonsense. That seems common? E.g., @gregkh on kernel reports and https://t.co/HPnGcBy11h. 23 Sept 2026
Show 17 more
- Were Ralph loops ever effective? They were popular, but when I did a comparison, they underperformed vs. a normal loop with context. There were all these theories about why Ralph loops were effective but AFAICT people just didn't run the comparison? 22 Sept 2026
- There's no point at which turning your brain off will work: 18 Sept 2026
- Amazon exec: you wouldn't notice [if someone blew up a datacenter]. I mean, we might be a bit upset, but you wouldn't notice! [laughs] Amazon after DCs blown up: After a thorough assessment, we have determined that we are unable to restore access to the resources and data ... 16 Sept 2026
- Interesting to see Yegge say he never successfully built anything with Gas Town. In https://t.co/YgJZCyLk9s, I mentioned not finding these ultra vibed orchestrators useful b/c reliability (w.r.t. completing tasks). Turns out the author of the most famous one had the same issue. 15 Sept 2026
- How well do agents use test/verification techniques? 7 Sept 2026
- This 2020 prediction that learning about security will be good for your career is looking pretty good right now. 4 Sept 2026
- How accurate have Ed Zitron's AI skeptic predictions been? 1 Sept 2026
- Bug Blindness: 29 Aug 2026
- There's no reason for software to be slow anymore: 21 Aug 2026
- The benchmarkpocalypse: 17 Aug 2026
- What programming languages are coding agents best at? 10 Aug 2026
- Exercises in evals and benchmarking: DeepSWE / Senior SWE-Bench, performance math, and cold weather tires 24 Jul 2026
- Agentic coding notes from Galapogos Island: 3 Jul 2026
- Exercises in benchmarking, evals, and experimental design, part 6: 8 Jun 2026
- In 10 years, will at least one of {Anthropic, OpenAI} be a public company valued at >= $1T in May 2026 dollars after adjusting for S&P 500 valuations (analogous to inflation adjustment, but using S&P 500 instead of CPI)? 29 May 2026
- Nowadays, I often see people say they have access to unlimited tokens. Usually for work, although in response to https://t.co/BuzKs79xk5, someone from Microsoft said they have unlimited tokens for personal projects. I wonder what this means in practice. Surely there's a limit? 25 May 2026
- Why are so many people so sure that the big AI providers are losing money on inference? It reminds me of the comments about how Uber can never make money. Their unit economics were fine and they were only losing money because they chose to do so on customer acquisition. 22 May 2026
Link verified 4 Sept 2026. Recent items update automatically from the channel.
Mastodon @danluu@mastodon.social Mirror of his posting.
Recent
- I had a non-anonymous human send some threats my way. I'm not exactly worried, but is there anything to do about this just in case? I guess there are data cleaning services that send data deletion requests, and I can make social media where I might've accidentally posted a photo that reveals a recent location private. Anything else one might generically do about this kind of thing? I have no doubt a motivated person could find me, so this would just be making things slightly less convenient. 2 Oct 2026
- "You should fix the error and try logging in again." 28 Sept 2026
- RE: https://mathstodon.xyz/@mjd/117326589463677193 I don't know mjd, but his writing demonstrates a combination of exceptionally clear thinking, reasonableness (in the Stanovich sense), strong knowledge of math, and the ability to explain difficult concepts simply. He seems like one of the rare people who has a deep understanding of abstract mathematical concepts who's grounded in reality enough to connect math to the real world. If I was hiring, I'd reach out to him to try to hire him, but since I'm not hiring, maybe you can hire him :-) 25 Sept 2026
Show 17 more
- Why didn't Anthropic do effective false positive rejection with their Mythos/Glasswing vuln reports? In https://danluu.com/ai-coding/#mythos, I mentioned a colleague finding that the reports we got were mostly nonsense. At the time, I wondered our reports were particularly bad, but as more people talk about their experiences, getting mostly nonsense reports seems to have been common? E.g., here's @gregkh on Linux kernel reports (h/t @KernelRecipes); see also https://www.vulncheck.com/blog/anthropic-glasswing-receipts. 23 Sept 2026
- Did Ralph loops ever actually work? They were popular for a while, but when I did a comparison, they underperformed vs. a normal loop. There were all these theories about why Ralph loops were effective ("you can invert the latent space of the model"), but AFAICT people just didn't run the comparison? 22 Sept 2026 danluu.com
- There's no point at which turning your brain off will work: 18 Sept 2026 danluu.com
- Amazon exec: "you wouldn't notice [if someone blew up a datacenter]. I mean, we might be a bit upset, but you wouldn't notice! [laughs]" Amazon official statement after DCs were blown up: "After a thorough assessment, we have determined that we are unable to restore access to the resources and data hosted exclusively in this Region." (yes, more than one DC was blown up, but, like other failures, of course DCs being blown up aren't uncorrelated events) 16 Sept 2026
- Interesting to see Steve Yegge say that he never successfully built anything with Gas Town (other than Gas Town). In https://danluu.com/ai-coding/, I mentioned not finding any of these super vibed orchestration frameworks useful because the reliability was too low (in terms of them actually getting agents to complete a non-trivial task). Turns out the author of the best known of these frameworks had the exact same issue. 15 Sept 2026
- I've been surprised to find that sometime in past six months or year, ChatGPT has generally started returning better results than Google for me. OpenAI probably has fewer employees than Google Search alone, and what fraction of OAI works on search? On the order of 1%? Arguably an unfair comparison because ChatGPT takes 20x longer to return a result, but Google doesn't have a "20x slower but don't return https://danluu.com/seo-spam/ knob" 10 Sept 2026
- How well do agents use test/verification techniques? 7 Sept 2026 danluu.com
- How accurate have Ed Zitron's AI skeptic predictions been? 1 Sept 2026 danluu.com
- Bug Blindness: 29 Aug 2026 danluu.com
- RE: https://mastodon.social/@danluu/111037093079443322 With the discussions of how bot traffic is flooding websites, I was surprised to find that my CDN is "only" using 3.5 TB/mo, up 3x in 3 years. Human traffic is probably down because I haven't been active for the last couple years, so maybe we can call that a 4x to 5x increase due to bots? Maybe my site just isn't very appealing to bots, but from the way people talk, I was expecting a higher ratio. I would've guessed >= 100x from how people are talking; 10x would be nothing in terms of load. 24 Aug 2026
- There's no reason for software to be slow anymore: 21 Aug 2026 danluu.com
- The benchmarkpocalypse: 17 Aug 2026 danluu.com
- What's the best programming language for coding agents? 10 Aug 2026 danluu.com
- Earlier this year, the FTC Chair said "you're going to have a hard time keeping up with the number of cases we're gonna be bringing" on privacy. I guess this is starting? See https://www.ftc.gov/system/files/ftc_gov/pdf/Hims-Complaint-Redacted-E-Filed.pdf. The bulk of the complaint is about standard scammy practices, but there's this section where the FTC is going after them because they promised to keep information private but use tracking pixels from various ad providers, which inherently share information about the user information. 2 Aug 2026
- After a civil engineering prof mention that the people in their department who work on water quality/treatment tend to drink out of the water cooler and not drink tap water, I tried getting a water filtration system. "Everyone" says Vancouver has really good water, so I didn't expect much, but the system's pre-filter catches a lot of green flecks (copper). The RDA of copper is < 1mg, which is on the order of these flecks, so it seems like I was ingesting many times the RDA in copper daily. 28 Jul 2026
- In another variant of https://danluu.com/learn-what/, I caught up with a former colleague who worked on automated theorem proving. It turns out he's had an interesting career doing all sorts of interesting stuff using the skills he developed by spending a decade writing/using theorem provers. At one point, he said, "if you use X like a theorem prover, it works really well", which surprised me to hear, but of course this is a highly generalizable skill just like compilers or benchmarking/evals. 25 Jul 2026
- Exercises in benchmarking and evals: performance math, winter tires, and DeepSWE / Senior SWE-Bench, 24 Jul 2026 danluu.com
Link verified 4 Sept 2026. Recent items update automatically from the channel.
Bluesky @danluu.com Where he posts now.
Recent
- Why didn't Anthropic do effective false positive rejection with their Mythos/Glasswing vuln reports? In danluu.com/ai-coding/#m..., I mentioned a colleague finding that the reports we got were mostly false. That seems common? E.g., gregkh on kernel reports and 23 Sept 2026 vulncheck.com
- Did Ralph loops ever work? They were popular for a while, but when I did a comparison, they underperformed. There were all these theories about why Ralph loops were effective ("invert the latent space of the model"), but AFAICT people just didn't run the comparison? 22 Sept 2026 danluu.com
- There's no point at which turning your brain off will work: 18 Sept 2026 danluu.com
Show 17 more
- Amazon exec: "you wouldn't notice [if someone blew up a datacenter]. I mean, we might be a bit upset, but you wouldn't notice! [laughs]" Amazon after DCs blown up: "After a thorough assessment, we have determined that we are unable to restore access to the resources and data hosted ...." 16 Sept 2026
- Interesting to see Steve Yegge say he never successfully built anything with Gas Town. I mentioned not finding these super vibed orchestrators useful because the reliability was too low (w.r.t. completing non-trivial tasks). Turns out the author of the most famous one had the exact same issue. 15 Sept 2026
- How well do agents use test/verification techniques? 7 Sept 2026 danluu.com
- How accurate have Ed Zitron's AI skeptic predictions been? 1 Sept 2026 danluu.com
- Bug Blindness: 29 Aug 2026 danluu.com
- There's no reason for software to be slow anymore: 21 Aug 2026 danluu.com
- The benchmarkpocalypse: 17 Aug 2026 danluu.com
- How do programming languages impact token efficiency and correctness? 10 Aug 2026 danluu.com
- Earlier this year, the FTC Chair said "you're going to have a hard time keeping up with the number of cases we're gonna be bringing" on privacy. It seems like this is starting; the FTC has filed a case against a company for promising privacy while running ad trackers on their site, which share info 2 Aug 2026
- Exercises in benchmarking and evals, part 7: performance napkin math, DeepSWE / Senior SWE-Bench, and winter tires 24 Jul 2026 danluu.com
- Agentic test processes, LLM benchmarks, and other notes on agentic coding from Galapagos Island: 4 Jul 2026 danluu.com
- A version of Missile Command for the Commodore 64 where the bottom of your screen is the game state in memory and missiles cause memory corruption: csdb.dk/release/?id=.... In the video below, a missile broke my controls and caused my cursor to get stuck moving down and to the left. 22 Nov 2024
- I find it interesting/surprising that Tesla topped the www.iseecars.com/most-dangero... fatalities per mile ranking from 2018-2022. Fatality rate is strongly negatively correlated with price and weight and Teslas are much more expensive and heavier than average. 17 Nov 2024
- A funny side effect of the crackdown on "AI" scraping is that I keep getting banned from sites for browsing too quickly. I barely use reddit anymore and I still managed to get IP banned for scraping (the error message indicated that I should get in touch with them if I want to do bulk accesses). 4 Nov 2024
- Steve Ballmer was an underrated CEO 28 Oct 2024 danluu.com
- A former Apple engineer discusses Google product culture: > My director wore an Apple Watch and had an iPhone ... my VP too. Nobody was expected to eat the dog food and so few did. This was crazy to me coming from Apple .... 13 Aug 2024
- On the 2011-2012 FTC antitrust investigation of Google: 27 May 2024 danluu.com
Link verified 4 Sept 2026. Recent items update automatically from the channel.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.