https://danluu.com/

xx/xxPatreon posts11/26Test and verification technique effectiveness with GPT-6 Astra10/26“We should be more ambitious with the software we build”09/26There's no point at which turning your brain off will work09/26How well do agents use test and verification techniques?09/26Ed Zitron's AI prediction track record08/26Bug blindness08/26There's no reason for software to be slow anymore08/26The benchmarkpocalypse08/26How does programming language affect token efficiency and correctness?07/26Bad benchmarks and evals: Senior SWE-Bench, napkin math, and winter tires07/26Agentic test processes, LLM benchmarks, and other notes on agentic coding from Galapagos Island10/24Steve Ballmer was an underrated CEO08/24How good can you be at Codenames without knowing any words?06/24A discussion of discussions on AI bias05/24What the FTC got wrong in the Google antitrust investigation03/24How web bloat impacts users with slow devices02/24Diseconomies of scale in fraud, spam, support, and moderation02/24Why it's impossible to agree on what's allowed01/24Notes on Cruise's pedestrian accident01/24Why do people post on [bad platform] instead of [good platform]?12/23How bad are search results? Let's compare Google, Bing, Marginalia, Kagi, Mwmbl, and ChatGPT09/22Futurist prediction methods and accuracy04/22In defense of simple architectures03/22Why is it so hard to buy things that work well?02/22Misidentifying talent02/22A decade of major cache incidents at Twitter02/22Cocktail party ideas12/21The container throttling problem12/21Some thoughts on writing12/21Some latency measurement pitfalls11/21Major errors on this blog (and their corrections)11/21Individuals matter11/21Culture matters10/21Willingness to look stupid10/21What to learn10/21Some reasons to work on productivity and velocity09/21The value of in-house expertise08/21Measurement, benchmarking, and data analysis are underrated12/20Against essential and accidental complexity06/20How do cars do in out-of-sample crash testing?05/20A simple way to get more value from tracing05/20A simple way to get more value from metrics03/20How (some) good corporate engineering blogs are written03/20The growth of command line options, 1979-Present02/20Suspicious discontinuities02/2095%-ile isn't that good01/20Algorithms interviews: theory vs. practice07/19Files are fraught with peril02/19Randomized trial on gender in Overwatch12/17Computer latency: 1977-201711/17How good are decisions? Evaluating decision quality in domains where evaluation is easy11/17How out of date are Android devices?11/17UI backwards compatibility10/17Filesystem error handling10/17Keyboard latency08/17Branch prediction08/17Sattolo's algorithm07/17Terminal latency06/17The widely cited studies on mouse vs. keyboard efficiency are completely bogus06/17Startup options v. cash02/17How web bloat impacts users with slow connections10/16HN comments are underrated10/16Programming book recommendations and anti-recommendations10/16Hiring and the market for lemons10/16I could do that in a weekend!09/16Is dev compensation bimodal?09/16How I learned to program08/16Notes on concurrency bugs04/16Some programming blogs to consider reading04/16Google SRE book03/16We only hire the trendiest01/16Sampling v. tracing01/16We saw some really bad Intel CPU bugs in 2015 and we should expect to see more in the future12/15Normalization of deviance12/15Big companies v. startups12/15Files are hard11/15Why use ECC?11/15What's worked in Computer Science: 1999 v. 201511/15Infinite disk10/15Why Intel added cache partitioning09/15Slowlock08/15Steve Yegge's prediction record08/15Reading postmortems05/15Slashdot and Sourceforge05/15The googlebot monopoly05/15A defense of boring languages05/15Advantages of monorepos05/15We used to build steel mills near cheap power. Now that's where we build datacenters03/15Reading citations is easier than most people think03/15Given that we spend little effort on testing, how should we test software?03/15What happens when you load a URL?03/15Goodhearting IQ, cholesterol, and tail latency02/15AI doesn't have to be very good to displace humans02/15CPU backdoors01/15Blog monetization01/15What's new in CPUs since the 80s?12/14A review of the Julia language12/14Integer overflow checking cost12/14Malloc tutorial12/14Markets, discrimination, and "lowering the bar"11/14TF-IDF linux commits11/14Everything is broken11/14Speeding up this site by 50x11/14How often is the build broken?11/14Literature review on the benefits of static types11/14CLWB and PCOMMIT11/14Caches: LRU v. random11/14Testing v. informal reasoning10/14Assembly v. intrinsics04/14Data-driven bug finding03/14Editing binaries03/14That bogus gender gap article03/14That time Oracle tried to have a professor fired for benchmarking their database02/14Why don't schools teach debugging?01/14Do programmers need math?01/14Data alignment and caches12/13PCA is not a panacea11/13Why hardware development is hard10/13How to discourage open source contributions10/13Randomize HN09/13Writing safe Verilog09/13Verilog is weird ↑ Public posts ↓Patreon posts06/26Exercises in benchmarking, evals, and experimental design, part 603/26Learning a bit about building game AIs01/26Exercises in benchmarking and experimental design, part 512/25Useless information about poker chips11/25How earthquake safe are Vancouver condos?04/25Exercises in benchmarking and experimental design, part 403/25The review-industrial complex (and the analyst newsletter industrial complex)12/24Exercises in benchmarking and experimental design, part 311/24Some comments on moving to Canada10/24Concussions and modern medicine09/24The utter dominance of U.S. tech07/24Exercises in benchmarking and experimental design, part 207/24What's wrong with Crockford's DEC64 type?06/24Exercises in benchmarking, part 1 of N06/24A lot of things have actually improved quite a bit, or how the monthly "everything sucks" rant is wrong (draft?)06/24The cost of housing construction in Vancouver05/24Greedy landlords, l