@thsottiaux Hey Tibo. I love sub-agents, but there's gotta be some indicator/badge on the exact model + reasoning effort being used in each sub-agent tab. I have workflows that use Sol as an orchestrator + Luna as a workhorse, but I can't confirm if it's even spawning the right models.
@thsottiaux Multi-project context for the Codex app. The CLI has `--add-dir` on startup, but what we really need is a dynamic `/add-dir` command like in Claude Code.
The Ralph Wiggum plugin for Claude Code has been getting a ton of attention lately.
Is it just me... or is it strange that Anthropic is shilling a plugin that's literally designed to consume as many tokens as necessary? 🤔
You're slicing your architecture wrong!
MVC has been a ubiquitous architectural pattern in software engineering—sometimes to the point of dogma.
In this article, we'll explore a better way to structure a codebase using vertically sliced architectures. 👇
https://t.co/A9pC9dyKgs
It is common wisdom in the programming world to discourage the use of global variables. I've recently come across a bug that humbled and reminded me about these pitfalls. Let me tell you a cautionary tale about global configuration objects.
https://t.co/8FwVesTQ7I
And finally, don't forget to star the GitHub repository! This is where I uploaded all of my code and results so that you can reproduce the data yourself and perhaps even extend my experiments. 🚀
{ 🧵 8/8 }
https://t.co/NXhk9Dcnfe
So what do we make of this data? Read more about my methodology, results, interpretations, and conclusions in the full article below.
{ 🧵 7/8 }
https://t.co/jSR21YfsNP
Quite impressively actually! Out of all the runtimes, Bun exhibited the best and most stable performance characteristics. But despite that, generators still performed the worst out of all the cases. That's 0/3 for me now. 😞
{ 🧵 6/8 }
I've always thought that #javascript generators were faster than chained array methods, but I've never measured if this is actually true. Let's talk about some of the key findings in my latest @ThePracticalDev article.
{ 🧵 1/8 }
https://t.co/jSR21YfsNP
Just to emphasize the similarities of the sample distributions for Node.js and Deno, look at the overlap between the histograms! Node.js and Deno are practically one and the same. So how does Bun fare?
{ 🧵 5/8 }
Wait, didn't we already see this plot? Yes, we did! Being built on top of the same V8 JavaScript engine, Node.js and Deno exhibited eerily identical performance characteristics. Though, that does mean I'm now 0/2 on the superiority of generators.
{ 🧵 4/8 }
After 2,048 warm-up iterations and 16,384 actual trials, the results are in! Much to my dismay and chagrin, I was not only wrong about generators, but it was also the worst performer among everyone in the Node.js trials. How about Deno, though?
{ 🧵 3/8 }
I measured four ways to implement a simple filter-then-map operation: (1) raw `for` loops, (2) `for...of` loops, (3) `filter` + `map`, and (4) generators. To be extra thorough, I ran these experiments on Node.js (V8), Deno (V8), and Bun (JavaScriptCore).
{ 🧵 2/8 }
@LizardOrman@tsoding Yes, that's true! I'm so glad for these new features, but the grammar is still inherently "right-leaning". Consider the nesting of impl > fn > match > code. That's already three levels of indentation without even mentioning control flow!