Mehmet Hakan Satman

Mehmet Hakan Satman

@jbytecode.bsky.social

Professor of Numerical Methods @ Istanbul University | Associate Editor @ Journal of Open Source Software (JOSS)

432 followers284 following257 posts

Mehmet Hakan Satman reposted

Our October newsletter is out 📬 🚀 Dyad 3.4: now a standalone desktop app 🧠 Same AI model: 0.899 in Dyad Agent vs 0.533 in Claude Code 📈 Julia is back in the TIOBE top 20 ⚡ Julia 1.13: ~30% faster precompilation juliahub.com/blog/october... #JuliaLang #SciML

juliahub.comOctober 2026 Newsletter: Dyad 3.4 and The Harness is Half the Model - Blog | JuliaHubExplore Dyad 3.4, physical AI benchmarks, Julia’s TIOBE top 20 return, lunar landing simulations, and upcoming JuliaHub webinars and workshops.

🟢JuliaConnectoR: An R package for calling Julia functions from within R "...The goal is to call functions from Julia packages directly as R functions. Julia functions imported via the JuliaConnectoR can accept and return R variables..." github.com/stefan-m-len... #JuliaLang

github.comGitHub - stefan-m-lenz/JuliaConnectoR: A functionally oriented interface for calling Julia from RA functionally oriented interface for calling Julia from R - stefan-m-lenz/JuliaConnectoR

🔴🟢🟣🔵 Do you miss tagged unions (Algebraic data types) and pattern matching when you code in Julia? MLStyle.jl package of Julia introduces parameterized enums and pattern matching as we do in OCaml, Scala, and F#. #JuliaLang

I've been looking for the most suitable GUI tool for Julia for a while now. I think the best approach is to use WebSockets from the HTTP package and connect the browser to Julia. (1/n) #JuliaLang

View full thread

I was looking for the best UI tool for Julia. HTTP.jl and WebSockets look like the best light-weight tool for communicating Julia and browser. This is my proof-of-concept app to get things together.

I've been looking for the most suitable GUI tool for Julia for a while now. I think the best approach is to use WebSockets from the HTTP package and connect the browser to Julia. (1/n) #JuliaLang

View full thread

I've been looking for the most suitable GUI tool for Julia for a while now. I think the best approach is to use WebSockets from the HTTP package and connect the browser to Julia. (1/n) #JuliaLang

So, partly as a proof of concept, I developed a tiny application. The app uses SQLite as a database and allows taking personal notes via an index.html file. We don't need a HTTP server, just opening the html file using Chrome is enough while server is active.

lts function of LinRegOutliers.jl is now faster than the R counterpart ltsreg of MASS when the regression data has 10000 observations and 10 parameters. Note that the R versiın is based on the native code whereas the Julia implementation is in pure Julia. github.com/jbytecode/Li... #JuliaLang

github.comGitHub - jbytecode/LinRegOutliers: Direct and robust methods for outlier detection in linear regressionDirect and robust methods for outlier detection in linear regression - jbytecode/LinRegOutliers

🟢A Julia package for creating probability distributions parameterised by moments (mean, variance, etc.) or quantiles, on standard or arbitrary supports. Builds on Distributions.jl. It makes this possible: d = make_dist(Normal, q1=-1, q3=1) #JuliaLang discourse.julialang.org/t/distributi...

discourse.julialang.orgDistributionsFactories.jlI’m happy to announce DistributionsFactories.jl, a package that solves a problem most of us have hit: you know the summary statistics you want — a mean and variance, a couple of quantiles, a mode — bu...

First, I define the data structures for the code I'm going to write (struct/class/enum/type). Then, I write a few initializers by hand. After that, I enter the prompts for the core functions as comments. I guess they call this the dinosaur style of using AI.

🟢 - "After six years, it’s time to release Pluto version 1.0! With this release, I want to celebrate all the progress over the past years, and to symbolise that Pluto is ready. I am proud of what we have achieved, and I hope you enjoy it!" #JuliaLang discourse.julialang.org/t/pluto-1-0-...

discourse.julialang.orgPluto 1.0 release!After six years, it’s time to release Pluto version 1.0! With this release, I want to celebrate all the progress over the past years, and to symbolise that Pluto is ready. I am proud of what we have a...

DuckDB is an excellent alternative to SQLite. It compressed my 114 MB historical price database down to just 8 MB while significantly speeding up query performance. Furthermore, it offers seamless integration with Julia via the DuckDB.jl package. #JuliaLang

DuckDB stores the data by columns, so column-wise queries are faster whereas SQLite stores the data by rows. It also zips the data so the size of databases is shorter. The REPL also has a more modern look and feel. SQLite has better speed-up in insert into type queries.