Installing collected packages: pytz, multipledispatch, tzdata, typing-extensions, toolz, threadpoolctl, six, setuptools, pygments, pillow, packaging, numpy, mdurl, kiwisolver, fonttools, filelock, cycler, cloudpickle, cachetools, scipy, python-dateutil, markdown-it-py, logical-unification, h5py, contourpy, rich, pandas, matplotlib, h5netcdf, cons, xarray, etuples, xarray-einstats, miniKanren, pytensor, arviz, pymc
Emphasis mine. I only wanted
PyMC and I've got miniKanren and logical-unification for free. Nice, but why, I wonder...https://www.youtube.com/watch?v=vMDHpPN_p08
Douglas Crockford gave a sorta first-hand overview of the (arrested) development of Programming Paradigms and was promoting the Actor Model as the way forward.
His only complaint about Erlang was that the addresses of the actors/processes are guessable, and thus do not provide the desired security level. Yeah, everything else about Erlang/Elixir is great. 😊
Douglas Crockford gave a sorta first-hand overview of the (arrested) development of Programming Paradigms and was promoting the Actor Model as the way forward.
His only complaint about Erlang was that the addresses of the actors/processes are guessable, and thus do not provide the desired security level. Yeah, everything else about Erlang/Elixir is great. 😊
YouTube
From Here To Lambda And Back Again - Douglas Crockford - RacketCon 2023
Renowned computer scientist Douglas Crockford delivers a thought-provoking keynote address at RacketCon 2023, tracing the evolution of programming language paradigms and presenting a compelling argument for the resurgence of the actor model, a paradigm that…
👍1
https://topos.institute/blog/2025-08-15-incremental-adhesive/
"Incremental query updating in adhesive categories"
And they freaking deliver on that promise!
That might not completely solve the problem of Differential Dataflow, especially in the case of SQL queries — I suspect the queries and/or updates don't have the right shape in these cases. But even solving incremental https://egraphs.org/ is a huge boon for a lot of domains.
"Incremental query updating in adhesive categories"
Working in the general setting of adhesive categories, we derive a practical algorithm for incrementally updating a query’s results with respect to small changes in the object being queried.
How can we unify and automate the intuition of the above examples, in the form of an algorithm that is a couple of lines long and self-evidently correct?
And they freaking deliver on that promise!
That might not completely solve the problem of Differential Dataflow, especially in the case of SQL queries — I suspect the queries and/or updates don't have the right shape in these cases. But even solving incremental https://egraphs.org/ is a huge boon for a lot of domains.
Topos Institute
Incremental query updating in adhesive categories – Topos Institute
Working in the general setting of adhesive categories, we derive a practical algorithm for incrementally updating a query’s results with respect to small changes in the object being queried.
❤3
https://dmkpress.com/catalog/computer/data/978-5-93700-425-3/
Роман Душкин скатился с Haskell в Prompt Engineering... O tempora, o mores!
Роман Душкин скатился с Haskell в Prompt Engineering... O tempora, o mores!
Dmkpress
Метакогнитивная промпт-инженерия
Купить книгу «Метакогнитивная промпт-инженерия», автора Душкин Р.В. в издательстве «ДМК Пресс». Выгодные цены в Москве, доставка. Заказать книги и учебники на официальном сайте издательства.
😁8😢3
A Web page says:
Get in touch with us to learn more about our mission and work, donate to our cause, or to become involved yourself. The best way to do this is by sending us an email:
info [at] topos [dot] institute
You can also find us on the following sites:
To what email should I write to reach them?
You should write to info [at] topos [dot] institute.
The page specifically instructs you to use that email address. The "[at]" and "[dot]" are used to prevent bots from automatically harvesting the email address. It means:
[at] should be replaced with the "@" symbol.
[dot] should be replaced with the "." symbol.
So, the full email address is info@topos.institute.
Answers Gemma3 12b running on my laptop. I don't think inserting
[at] and [dot] into email address prevent automatic harvesting anymore...👎1😢1
https://lawrencecpaulson.github.io/2025/09/20/Wrong.html
Among the few advantages of attaining the dizzy age of 70 is the ability to look back on half a century. Things were great back in 1975. ... Great progress had been made in Artificial Intelligence, thanks to a relentless focus on symbol processing as opposed to discredited, useless neural networks. Many thought that automatic theorem provers could lead the way to what we now call AGI. Also, watches did not have operating systems. People knew that clouds were real and that vaccines worked. Well, a lot can happen in 50 years.
🔥2
https://doom-captcha.vercel.app/
DAAC: DOOM-as-a-Captcha 😏
I long forgot how hard it is without mouse look and strafe...
DAAC: DOOM-as-a-Captcha 😏
I long forgot how hard it is without mouse look and strafe...
DOOM® CAPTCHA
Prove you're human by playing DOOM
https://causalai-book.net/
Elias Bareinboim published a "draft" (comprising 1200 pages!) of his "Causal Artificial Intelligence" #book for #free.
Bareinboim spent years connecting Causal and Counterfactual Reasoning and #causalinference following Judea Pearl with Machine Learning and Reinforcement Learning in particular. So the book tells it all: from the fundamentals of Structured Causal Models (with infamous backdoors, frontdoors and d-separation) to Causal Reinforcement Learning to Neural Causal Models and "Causal GenAI" and beyond.
If you understandably don't have the time to read through the volume front to back, Bareinboim published most of the slides for his lectures following the book, and the rest are coming.
Elias Bareinboim published a "draft" (comprising 1200 pages!) of his "Causal Artificial Intelligence" #book for #free.
Bareinboim spent years connecting Causal and Counterfactual Reasoning and #causalinference following Judea Pearl with Machine Learning and Reinforcement Learning in particular. So the book tells it all: from the fundamentals of Structured Causal Models (with infamous backdoors, frontdoors and d-separation) to Causal Reinforcement Learning to Neural Causal Models and "Causal GenAI" and beyond.
If you understandably don't have the time to read through the volume front to back, Bareinboim published most of the slides for his lectures following the book, and the rest are coming.
causalai-book.net
Causal Artificial Intelligence Book
👍2🔥1
https://www.hillelwayne.com/post/algdt-history/
Turns out John McCarthy invented Algebraic Data Types and described their semantics in terms of finite sets. Also he (incidentally) was using the term spaces for the types. 😏
Turns out John McCarthy invented Algebraic Data Types and described their semantics in terms of finite sets. Also he (incidentally) was using the term spaces for the types. 😏
Hillel Wayne
A Very Early History of Algebraic Data Types
Been quiet around here! I’ve been putting almost all of my writing time into Logic for Programmers and my whole brain is book-shaped. Trust me, you do not want to read my 2000-word rant on Sphinx post-build LaTeX customization. But I spent the past week in…
🔥1
https://con.racket-lang.org/
October 4-5
Some very interesting talks, as per usual.
Online participation is possible, but costs like USD $10. Internet isn't free, after all. 😏
(fifteenth RacketCon)October 4-5
Some very interesting talks, as per usual.
Online participation is possible, but costs like USD $10. Internet isn't free, after all. 😏
AlexTCH
https://con.racket-lang.org/ (fifteenth RacketCon) October 4-5 Some very interesting talks, as per usual. Online participation is possible, but costs like USD $10. Internet isn't free, after all. 😏
Actually, that last bit is wrong:
The stream will be free to watch, but you can buy a ticket to support the conf.
The live stream is publicly available, but buying this ticket helps pay for the live stream and ensure its availability for the entire community.
The stream will be free to watch, but you can buy a ticket to support the conf.
🔥3
This is old, but still great news.
https://avehtari.github.io/ActiveStatistics/
Active Statistics by Andrew Gelman and Aki Vehtari
A #free introductory #book to (Applied) Statistics from some of the best teachers and practitioners of the craft alive. And that's not all what's great about it.
The book is unusual in that it's aimed at (university) teachers. And thus the material is split between two semester and weekly classes, while the first part explains the pedagogy of the book: the "flipped classroom" approach, the role of computation and simulation, setting up assessment, and so on.
The classes themselves are centered around "stories", which are supported by relevant activities for the students, computer demonstrations and open-ended questions to discuss.
Gelman was deeply thinking about what makes a good educational "story", and gave a couple of talks summarizing his conclusions and approach. Thus the stories in the book are far from random and purely entertaining anecdotes. They capture the essence of the topic and provide the bridge between more abstract statistical notions and techniques and the real-world applications.
While invaluable for teachers, the book is highly applicable for self-study or study group, as soon as you become your own teacher and can leverage the structure, examples and exercises.
https://avehtari.github.io/ActiveStatistics/
Active Statistics by Andrew Gelman and Aki Vehtari
A #free introductory #book to (Applied) Statistics from some of the best teachers and practitioners of the craft alive. And that's not all what's great about it.
The book is unusual in that it's aimed at (university) teachers. And thus the material is split between two semester and weekly classes, while the first part explains the pedagogy of the book: the "flipped classroom" approach, the role of computation and simulation, setting up assessment, and so on.
The classes themselves are centered around "stories", which are supported by relevant activities for the students, computer demonstrations and open-ended questions to discuss.
Gelman was deeply thinking about what makes a good educational "story", and gave a couple of talks summarizing his conclusions and approach. Thus the stories in the book are far from random and purely entertaining anecdotes. They capture the essence of the topic and provide the bridge between more abstract statistical notions and techniques and the real-world applications.
While invaluable for teachers, the book is highly applicable for self-study or study group, as soon as you become your own teacher and can leverage the structure, examples and exercises.
❤1
https://arstechnica.com/science/2025/09/meet-the-2025-ig-nobel-prize-winners/
Let's celebrate the victory march of Science!
Let's celebrate the victory march of Science!
https://www.youtube.com/watch?v=PnQ0eQrKF4g
Curious how our activities exert evolutionary pressure over various animal species, which gets recorded into their genes. Well, any activity on the planet exerts evolutionary pressure, that's the whole point. But humans act on unprecedentedly short time scales, and the nature notices (and adapts).
Curious how our activities exert evolutionary pressure over various animal species, which gets recorded into their genes. Well, any activity on the planet exerts evolutionary pressure, that's the whole point. But humans act on unprecedentedly short time scales, and the nature notices (and adapts).
YouTube
Animals Are Evolving to Survive in the Human World But Often in a Weird Way
Support this channel on Patreon to help me make this a full time job: https://www.patreon.com/whatdamath (Unreleased videos, extra footage, DMs, no ads)
Alternatively, PayPal donations can be sent here: http://paypal.me/whatdamath
Get a Wonderful Person Tee:…
Alternatively, PayPal donations can be sent here: http://paypal.me/whatdamath
Get a Wonderful Person Tee:…
❤1
https://www.answer.ai/posts/2025-06-05-readbench.html
The folks designed and implemented the
The results are kinda as expected, but still show that you can't simply pile all of your scanned documents into a DB and expect a VLM to efficiently sift through them all and extract necessary information.
The folks designed and implemented the
ReadBench benchmark for testing Vision-Language Models (VLMs) and other Multimodal LLMs essentially on the OCR tasks.The results are kinda as expected, but still show that you can't simply pile all of your scanned documents into a DB and expect a VLM to efficiently sift through them all and extract necessary information.
Answer.AI
TIL: Vision-Language Models Read Worse (or Better) Than You Think – Answer.AI
Introducing ReadBench, a straightforward way to see how well your favorite Vision-Language Models read text-rich images.
https://julialang.org/blog/2025/10/julia-1.12-highlights/
Julia 1.12 is out. This one largely continues the long-standing theme of improvements to compilation and packaging.
Compilation improvements include the option to build the Julia itself with LLVM BOLT (as well as LTO and PGO), and precompilation times tracing.
For building self-contained binary images there's now an expreimental option to
Other packaging improvements include support for workspaces comprising several closely related projects (like a library + tests + benchmarks + docs + examples) and dedicated (CLI) apps.
Julia 1.12 is out. This one largely continues the long-standing theme of improvements to compilation and packaging.
Compilation improvements include the option to build the Julia itself with LLVM BOLT (as well as LTO and PGO), and precompilation times tracing.
For building self-contained binary images there's now an expreimental option to
--trim all statically unreachable code, which yields way smaller binaries, but (expectedly) breaks on dynamic dispatches and other dynamic features. Looks very similar to AOT compilation for Java and C# which also breaks on reflection and similar dynamic features.Other packaging improvements include support for workspaces comprising several closely related projects (like a library + tests + benchmarks + docs + examples) and dedicated (CLI) apps.
julialang.org
Julia 1.12 Highlights
Highlights of the Julia 1.12 release.
🔥1
This year’s London Mathematical Society (LMS) / British Computer Society -- Formal Aspects of Computing Science (BCS-FACS) Evening Seminar will feature Jeremy Avigad as the distinguished speaker. Registration is free but required in advance.
https://www.lms.ac.uk/events/lms-bcs-facs-seminar-jeremy-avigad
Date: 6 November 2025
Time: 19:00 (UK time)
Format: Online via Zoom
Talk title: Mathematics in the Age of AI
Abstract:
New technologies for reasoning and discovery are bound to have a profound effect on mathematical practice. Proof assistants are already changing the nature of collaboration, communication, and curation of mathematical knowledge. Automated reasoning tools are used to find mathematical objects with specified properties or rule out their existence, and to decide or verify mathematical claims. Machine learning and neural methods can discover patterns in mathematical data, explore complex mathematical spaces, and generate mathematical objects of interest. Neurosymbolic theorem provers, now capable of solving the most challenging competition problems, combine aspects of all of these technologies.
It is helpful to keep in mind that the phrase "AI for mathematics" encompasses several distinct technologies that overlap and interact in interesting ways. In this talk, I will survey the landscape, describe a few landmark applications to mathematics, and encourage you to join me in thinking about how mathematicians and computer scientists can collaborate to guide mathematics through this era of technological change.
The talk shall be uploaded to YouTube as well as the previous ones:
https://www.youtube.com/channel/UCvp7jsbXjx2k8sGEkdtWCAw
https://www.lms.ac.uk/events/lms-bcs-facs-seminar-jeremy-avigad
Date: 6 November 2025
Time: 19:00 (UK time)
Format: Online via Zoom
Talk title: Mathematics in the Age of AI
Abstract:
New technologies for reasoning and discovery are bound to have a profound effect on mathematical practice. Proof assistants are already changing the nature of collaboration, communication, and curation of mathematical knowledge. Automated reasoning tools are used to find mathematical objects with specified properties or rule out their existence, and to decide or verify mathematical claims. Machine learning and neural methods can discover patterns in mathematical data, explore complex mathematical spaces, and generate mathematical objects of interest. Neurosymbolic theorem provers, now capable of solving the most challenging competition problems, combine aspects of all of these technologies.
It is helpful to keep in mind that the phrase "AI for mathematics" encompasses several distinct technologies that overlap and interact in interesting ways. In this talk, I will survey the landscape, describe a few landmark applications to mathematics, and encourage you to join me in thinking about how mathematicians and computer scientists can collaborate to guide mathematics through this era of technological change.
The talk shall be uploaded to YouTube as well as the previous ones:
https://www.youtube.com/channel/UCvp7jsbXjx2k8sGEkdtWCAw
👍1