If you found R the language and the libraries quite a bit clunky, crufty and somewhat annoying to make what you want, now you can have much better developer experience and output quality thanks to Uri Simonsohn and coding AI:
https://datacolada.org/132
https://datacolada.org/132
Data Colada
[132] statuser: R in user-friendly mode - Data Colada
t.test(), the R function for running t-tests, is disconcertingly imperfect.
https://www.kickstarter.com/projects/glnc/scp-gallionic
Another SCP-themed short film on Kickstarter. This time animated in the style of Anime.
Watch the trailer, it's really good. "Netflix-quality" in a good way.
Formally, the project is already funded, but they have a lot of very desirable stretch goals. The production timeline is long, which means they probably know what they're doing.
Another SCP-themed short film on Kickstarter. This time animated in the style of Anime.
Watch the trailer, it's really good. "Netflix-quality" in a good way.
Formally, the project is already funded, but they have a lot of very desirable stretch goals. The production timeline is long, which means they probably know what they're doing.
Kickstarter
SCP:GALLIONIC
An animated short set in the SCP universe following a breach and the effects it has on two siblings working within.
👍3
Modern Web Front-End Development time breakdown:
— setting up the project with package managers, bundlers and frameworks: 10%
— writing the code: 5%
— fighting the fucking CORS: 85%
— setting up the project with package managers, bundlers and frameworks: 10%
— writing the code: 5%
— fighting the fucking CORS: 85%
😁7👍1💯1
Research (Maths, Physics, Chemistry, Sociology, History, etc) is PvE.
Beaurocracy (programs, grants, management, etc) is PvP.
Beaurocracy (programs, grants, management, etc) is PvP.
💯4🔥1🤔1
https://www-cs-faculty.stanford.edu/~knuth/papers/claude-cycles.pdf
Donald Knuth discovers Vibe-coding and even Vibe-(paper)-proving through his friend Filip Stappers and his friend Claude. 😁
Actually, it's quite a story, and a pleasure to read. The Knuth's comments are deep and enlightening.
Donald Knuth discovers Vibe-coding and even Vibe-(paper)-proving through his friend Filip Stappers and his friend Claude. 😁
Actually, it's quite a story, and a pleasure to read. The Knuth's comments are deep and enlightening.
❤3
Two more somewhat recent blog posts about "vibe-coding" that kinda go along nicely:
https://www.anthropic.com/research/AI-assistance-coding-skills
and
https://www.modular.com/blog/the-claude-c-compiler-what-it-reveals-about-the-future-of-software
The first one is about a N=52 study from Anthropic on the question "how much knowledge vibe-coders accumulate about what they're doing?" The answer, like usual, "it depends". In this case, it depends on how you're using an AI assistant, and what questions you ask. One can both learn more from AI compared to no AI, or learn nothing at all — the choice is yours.
The second one is a pretty extensive commentary from Chris Lattner of LLVM and Clang fame on the (in)famous "Claude C Compiler" from Anthropic again. As someone living and breathing compilers pretty much his whole career, Lattner has some deep insights into the significance and challenges of compilers in general, and the "Claude C Compiler" in particular. Sure enough he also have biases. As a leader of an AI company Lattner cannot say "AI is bad", so his vision of "the evolving role of software engineers" might be overly optimistic, but he has good points nevertheless.
Which kinda circles back to the first post: it's not AI in itself, but the way one uses it that matters. Which approach will "win"? We'll see soon enough.
https://www.anthropic.com/research/AI-assistance-coding-skills
and
https://www.modular.com/blog/the-claude-c-compiler-what-it-reveals-about-the-future-of-software
The first one is about a N=52 study from Anthropic on the question "how much knowledge vibe-coders accumulate about what they're doing?" The answer, like usual, "it depends". In this case, it depends on how you're using an AI assistant, and what questions you ask. One can both learn more from AI compared to no AI, or learn nothing at all — the choice is yours.
The second one is a pretty extensive commentary from Chris Lattner of LLVM and Clang fame on the (in)famous "Claude C Compiler" from Anthropic again. As someone living and breathing compilers pretty much his whole career, Lattner has some deep insights into the significance and challenges of compilers in general, and the "Claude C Compiler" in particular. Sure enough he also have biases. As a leader of an AI company Lattner cannot say "AI is bad", so his vision of "the evolving role of software engineers" might be overly optimistic, but he has good points nevertheless.
Which kinda circles back to the first post: it's not AI in itself, but the way one uses it that matters. Which approach will "win"? We'll see soon enough.
Anthropic
How AI assistance impacts the formation of coding skills
Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
https://sockpuppet.org/blog/2026/03/30/vulnerability-research-is-cooked/
"Interesting" times for Computer Security research and practice. Could be even more "interesting" for us, consumers. "90's Windows security" vibes for those old enough to remember.
"Interesting" times for Computer Security research and practice. Could be even more "interesting" for us, consumers. "90's Windows security" vibes for those old enough to remember.
sockpuppet.org
Vulnerability Research Is Cooked
😁3
Некоторое время назад наше любимое издательство ДМК Пресс выпустило книжку для «настоящих компиляторщиков»:
https://dmkpress.com/catalog/computer/programming/978-5-93700-432-1/
«Разработка компиляторов на основе формы SSA» — в девичестве та самая «SSA-based Compiler Design» aka «The SSA Book».
Книга содержит 24 главы (на 400 страницах), покрывающих весь жизненный цикл SSA от построения, через алгоритмы анализа и преобразования, и до распределения регистров и выбора инструкций. За бортом остались только синтаксический разбор исходного языка и множество алгоритмов оптимизации — обе темы подробно раскрываются в других книгах.
Тем не менее, это не учебник по разработке компилятора, скорее справочник по теоретическим и практическим вопросам использования SSA. Авторы рассматривают разные варианты формы SSA, разные представления одних и тех же форм, разные алгоритмы построения и анализа.
Но в отличие от сухого справочника, книжка хорошо написана и легко читается для отвыкших от художественной литературы. Впечатление немного портит «упрощённый» перевод, например, базовые блоки называются «простыми», а «анализ живучести», видимо, стал стандартным термином в ДМК.
Как бы то ни было, издание «единственное и неповторимое» — другого просто не будет, так что...
https://dmkpress.com/catalog/computer/programming/978-5-93700-432-1/
«Разработка компиляторов на основе формы SSA» — в девичестве та самая «SSA-based Compiler Design» aka «The SSA Book».
Книга содержит 24 главы (на 400 страницах), покрывающих весь жизненный цикл SSA от построения, через алгоритмы анализа и преобразования, и до распределения регистров и выбора инструкций. За бортом остались только синтаксический разбор исходного языка и множество алгоритмов оптимизации — обе темы подробно раскрываются в других книгах.
Тем не менее, это не учебник по разработке компилятора, скорее справочник по теоретическим и практическим вопросам использования SSA. Авторы рассматривают разные варианты формы SSA, разные представления одних и тех же форм, разные алгоритмы построения и анализа.
Но в отличие от сухого справочника, книжка хорошо написана и легко читается для отвыкших от художественной литературы. Впечатление немного портит «упрощённый» перевод, например, базовые блоки называются «простыми», а «анализ живучести», видимо, стал стандартным термином в ДМК.
Как бы то ни было, издание «единственное и неповторимое» — другого просто не будет, так что...
🔥4
https://plr.csail.mit.edu/
Apparently, MIT runs an annual seminar where they discuss with authors some old and new programming-related papers. This year's selection covers a broad range of topics from (black-box) testing of shader compilers' back-ends to empirical study of «prompt engineering»; from a decade-old paper on database query optimization to optimizing Datalog for GPGPUs to compiling sequent calculus; from program synthesis to an interactive debugger for Rust traits.
The seminar is open for virtual attendees too on Friday May 8th, 2026.
Apparently, MIT runs an annual seminar where they discuss with authors some old and new programming-related papers. This year's selection covers a broad range of topics from (black-box) testing of shader compilers' back-ends to empirical study of «prompt engineering»; from a decade-old paper on database query optimization to optimizing Datalog for GPGPUs to compiling sequent calculus; from program synthesis to an interactive debugger for Rust traits.
The seminar is open for virtual attendees too on Friday May 8th, 2026.
🔥1
https://elixir-lang.org/blog/2026/01/09/type-inference-of-all-and-next-15/
For about five years Elixir contributors have been working on integrating a type system into the language and tooling. Starting with the development of a pretty novel and unorthodox type system.
I'm fascinated by several things:
— they still haven't dropped the project. Despite slow progress and obstacles in performance and usability they systematically overcome.
— they are deeply committed to the soundness of their gradual type system, which is tricky (and to my eyes they sometimes confuse soundness and completeness in the blogs at least)
— they put end-user (developer) usability front and center, in terms of speed and responsiveness, type inference to reduce annotation burden, meaningful error messages, etc.
Towards that goals the post references their work on
https://elixir-lang.org/blog/2023/09/20/strong-arrows-gradual-typing/
and the development of
https://elixir-lang.org/blog/2025/12/02/lazier-bdds-for-set-theoretic-types/
which are cool in their own right.
For about five years Elixir contributors have been working on integrating a type system into the language and tooling. Starting with the development of a pretty novel and unorthodox type system.
I'm fascinated by several things:
— they still haven't dropped the project. Despite slow progress and obstacles in performance and usability they systematically overcome.
— they are deeply committed to the soundness of their gradual type system, which is tricky (and to my eyes they sometimes confuse soundness and completeness in the blogs at least)
— they put end-user (developer) usability front and center, in terms of speed and responsiveness, type inference to reduce annotation burden, meaningful error messages, etc.
Towards that goals the post references their work on
https://elixir-lang.org/blog/2023/09/20/strong-arrows-gradual-typing/
and the development of
https://elixir-lang.org/blog/2025/12/02/lazier-bdds-for-set-theoretic-types/
which are cool in their own right.
The Elixir programming language
Type inference of all constructs and the next 15 months
Today we celebrate 15 years since Elixir’s first commit! To mark the occasion, we are glad to announce the first release candidate for Elixir v1.20, which performs type inference of all language constructs, with increasing precision.
👍2💩1
https://www.pathsensitive.com/2022/03/abstraction-not-what-you-think-it-is.html
Abstractions are not what they seem. Just like owls.
Cool post. It mentions λ-calculus, antiunification (and Gordon Plotkin), abstract interpretation, and even contains a primer on the fundamental theorem of homomorphisms from Gerry Sussman himself.
I'd only like to add that we can represent abstractions themselves in (some) programming languages, as long as we can reify abstract domains and abstraction mappings in the form of language constructs. Different constructs in different instances, but Type Classes indeed show up rather often.
Abstractions are not what they seem. Just like owls.
Cool post. It mentions λ-calculus, antiunification (and Gordon Plotkin), abstract interpretation, and even contains a primer on the fundamental theorem of homomorphisms from Gerry Sussman himself.
I'd only like to add that we can represent abstractions themselves in (some) programming languages, as long as we can reify abstract domains and abstraction mappings in the form of language constructs. Different constructs in different instances, but Type Classes indeed show up rather often.
Pathsensitive
Abstraction: Not What You Think It Is
“Interfaces are abstractions” — Olaf Thielke , the "Code Coach" “Interfaces are not abstractions” — Mark Seeman , author of Code that Fit...
BREAKING NEWS!
Lawrence C. Paulson of the Isabelle/HOL fame now also has a YouTube channel:
https://www.youtube.com/@lawrpaulson
Lawrence C. Paulson of the Isabelle/HOL fame now also has a YouTube channel:
https://www.youtube.com/@lawrpaulson
🎉6