Ruslan Shevchenko retweetledi
Ruslan Shevchenko
6.9K posts
Ruslan Shevchenko
@rssh1
github: https://t.co/LcOh2UgLzO columsn on dou: https://t.co/HWcu7MGR5u… blog: https://t.co/YqobW9Fn2d
N 49°34' 0'' / E 32°13' 0'' Katılım Haziran 2010
939 Takip Edilen790 Takipçiler
@clarkalesna @wolf31o2 @Quantumplation @atlanter @scalus // will submit the path when I am near the computer.
English
@clarkalesna @wolf31o2 @Quantumplation @atlanter Actually, coop-benchmark is invalid: it uses a case on a boolean, which will work in the upcoming Plutus release. @Scalus implements this under the future flag, which is not enabled by default.
English

I’m single handedly improving Cardano smart contract performance by pressuring everyone 🤣😅🤣
@wolf31o2 @Quantumplation update live
@atlanter put a category for vms not able to run all benchmarks for now, let me know once you have an update that can run all the benchmark 🙌

English
Ruslan Shevchenko retweetledi
@Pitometsu @Scalus3 @lantr_io Recently partial evaluation of uplc was added, it's close to supercompilation, but not yet exactly the same.
English

Does @Scalus3 use a supercompilation for the optimization? @lantr_io stackoverflow.com/a/10888771/319…
English
Ruslan Shevchenko retweetledi

Популярність мов програмування за останні 5 років в українському ІТ
TypeScript продовжує стабільно зростати. Частка Python після стрімкого зростання трохи зменшилася. JavaScript знижується. Частка C# і Java повільно, але стабільно зменшується.
Рейтинг мов програмування 👉 dou.ua/goto/mHLa

Українська
Ruslan Shevchenko retweetledi
Ruslan Shevchenko retweetledi
Ruslan Shevchenko retweetledi

Looking for an internship this summer/fall? The JetBrains Internship Program 2026 is now open!
If Scala is your thing, we might have something just for you: a project focused on improving error highlighting in the IntelliJ Scala Plugin.
Learn more: internship.jetbrains.com/projects/1756
English
about monad for reinforcement learning: github.com/rssh/notes/blo…
English
dotty-cps-async 1.3 with tailRecM. Announce on reddit:
reddit.com/r/scala/commen…
English

I used to use a single Claude Code session. It made so many mistakes that just preventing it from destroying everything was a nightmare
Nowadays, with just a few commands/skills and prompts, Opus 4.5 can show some reliability. I need to babysit much less now. Working on multiple parallel sessions doing research, debugging issues, exploring designs, evaluating optimizations, etc feels like a new super power 🤯
I know there's a lot of skeptics out there. And yes, many are correct. LLMs are likely not reliable enough to replace developers and there's no good evidence they can truly extrapolate outside of their training set. True tech evolution is inherently creative work that LLMs struggle at
That said, unless we end up with LLMs that are too expensive (pretty much all use these days is subsidized!), I don't see how this tech won't forever change the field
Nice times 😎
English
Ruslan Shevchenko retweetledi
@tribbloid Oz initially used this model (all independent tasks run with max concurrency by default), but later changed it to implicit thread creation for blocking tasks. Now the proportions of resources for switching/creating are changed, but still non-trivial
English

@rssh1 yeah turns out this rewriting is quite verifiable with several checkpoints in between, the bottleneck must be elsewhere
English

who is working on this?
AI agents are good at wielding simple, sequential code that are slow, but it shouldn't be hard to use a CPS compiler to accelerate it for parallel execution or multi-agent orchestration
that's a lot of performance gain
Erik Meijer@headinthebox
Here is what's cooking: Tx is a direct-style terminal UI framework for Kotlin - the counterpart to Rx (Reactive Extensions). Where Rx uses observables and subscribers, Tx uses suspend functions and sequential code. No callbacks, no event handlers, just straight-line code that suspends on user input.
English









