I have been beta testing this new update for some time, and I really can't wait for the update to drop for the rest of the players. It looks absolutely gorgeous compared to the state when I first found it – https://www.youtube.com/watch?v=JTRBK6Nv7oA
Hello PL enthusiasts!
Lately I have been thinking about how to best introduce a language to a newcomer... And I kept coming back to http://clojurescriptkoans.com – try it if you haven't, it's awesome. It teaches you Clojure(Script) via series of progressively harder challenges. The instructions might sometimes be cryptic – as Zen koans often are, but once you understand, they become obvious.
While that served as my inspiration, I ended up somewhere else. The gates are open, so you can passively enjoy the ride, or you can actively participate if you feel like it. Please let me know what do you think.
Hello,
after 6 months of conceiving this idea, I finally got significant inline white-space working in Fluent. Let me explain...
Fluent has three strict syntax rules:
For example: 1 + 2 * 3 - 4 / 5 is evaluated as (((1 + 2) * 3) - 4) / 5. If you would want to emulate operator precedence, you'd have to use parens to express intent: 1 + (2 * 3) - (4 / 5). With significant inline white-space, you can now express intent by "gluing" parts together – 1 + 2*3 - 4/5 without using parens.
A second rule of significant inline white-space is "unbalanced gluing". This is especially handy when you need to use binding/assignment, which is just another operator and left-to-right flow still applies. While x : 1 is okay, x : 1 + 2 is not, because it is parsed as (x : 1) + 2, which is obviously wrong. Normally you'd have to enclose the assignment value in parens: x : (1 + 2) , but this becomes very annoying. By gluing the operator to the left argument, you create a long right scope, so x: 1 + 2 gets parsed as x : (1 + 2), which is exactly what you wanted.
With these two simple rules, left-to-right no-precedence flow became super ergonomic.
---
Fluent is a tiny lang for differentiable tensors and reactive programming. More at project page and live REPL. It originated in 2021 as a language for the New Kind of Paper project, which aims to fulfill the original vision of APL – a handwritten & unambiguous notation for executable math.
Hey, this is just a small appreciation post to whomever FUCKING NAILED the stereoscopic anti-aliasing on the visionOS team. From window edges to text, everything is so damn crisp. Magic!
Hello, does anybody know if Havnby FlatCore for Y Juniper works in 3 Highland, please?
Hello World!
For past 2 weeks I have been exploring how software agents can be paired with Smalltalk. This demo video is a result of my findings. (Please enable subtitles for more coherent language. Sorry about the filler words, I'll do better narration job next time.)
TL;DW: Blueprint is a rework of Cuis desktop experience and has been done purely by code agent directed in natural language by me. No Smalltalk code was written by hand. I am not stopping, I have too much fun. :)
If you have any questions, I am happy to answer them.
Yesterday I expressed a hunch about Smalltalk having a huge advantage for agentic coding. Some pushback forced me to get my hands dirty and try it out myself. And I am happy to report that my hunch was spot on.
I tried these 3 tests:
All of them were successfully completed. Claude inside Smalltalk is a badass. Now onto some Cuis!
Smalltalk has an extreme advantage when it comes to productivity of a single person. Having ability to touch any part of the system is an extreme leverage. With code gen revolution happening this is even more obvious. Yet, I have to see a breath-taking demo of code-gen in action for Smalltalk. Please point me to some if you have. Thank you.