Awesome to see the progress made by Alex, highly recommend watching the linked talk for people interested in any of Ruby, (AoT) compilers, and MLIR (which I haven't seen used too much outside of the ML space, personally)
Just-in-time compilation of Ruby allowing you to elide a lot of the overhead of dynamic language features + executing optimized machine code instead of running in the VM / bytecode interpreter.
For example, doing some loop unrolling for a piece of code with a known & small-enough fixed-size iteration. As another example, doing away with some dynamic dispatch / method lookup for a call site, or inlining methods - especially handy given Ruby's first class support for dynamic code generation, execution, redefinition (monkey patching).
> In particular, YJIT is now able to better handle calls with splats as well as optional parameters, it’s able to compile exception handlers, and it can handle megamorphic call sites and instance variable accesses without falling back to the interpreter.
> We’ve also implemented specialized inlined primitives for certain core method calls such as Integer#!=, String#!=, Kernel#block_given?, Kernel#is_a?, Kernel#instance_of?, Module#===, and more. It also inlines trivial Ruby methods that only return a constant value such as #blank? and specialized #present? from Rails. These can now be used without needing to perform expensive method calls in most cases.
I recommend trying out Quiche Browser on iOS, or Orion from Kagi- both will support Kagi as a search option. Spotlight search still goes to Google, but I find that acceptable.
I had similar thoughts about a year ago. My experience is that the results are that good, though sometimes there are changes under the hood which lead to poor or limited results- it is a beta project, after all. I used the free trial, following which I stayed on the monthly sub before paying a lump sum of $100 for the next 10 months. I've enjoyed my correspondence with the CEO as well. YMMV
Happy to see this here; I found it to be a fun language to prototype reasonably performant things in + for learning about certain data structures. I expect to see the comments talking about the lack of docs, language server struggles, and compilation times (the whole stdlib and program is compiled from scratch every time a program is compiled/run, IIRC). The language is slowly growing and I encourage people to stick with its onboarding shortcomings and giving it a shot for their next project.
> I am so bad at iterating over dataframes! It always feels horrible and slow. While doing this though, I discovered that using df.to_dict('records') and then iterating over the resulting dictionary is almost 100x faster than using the pandas built-in iteration tools like df.itertuples() or df.iterrows()!
That's really surprising to hear, any context on why this is? Very fun read BTW, my friends and I have joked about making something similar for our DMs (nicknamed MattGPT) and giving "them" topics to discuss + observing what they come up with.
Depends on your use case and financials. I use it extensively for work as a computer scientist + developer, and have expensed it in the past. I initially found the free plan of ~30 searches / month pretty good for a 'sudo search', when Google and others weren't quite cutting it and I felt that there could be better results.
I agree about the efficiency and wouldn't expect anyone to know that on a thread, off of a handle alone. However, I see it in a more positive light- based on John's other comments in the thread, he's made the time to stay active in communities like this one even as the CTO and followed up with folks internally to understand how an oversight like this could've happened.
Great stuff, I went down the rabbit hole of building something similar for synthesizing flash cards + Q/A pairs from textbook PDFs about a year ago, and I would also emphasize that PDF search is a janky nightmare to get within the ballpark of usability :')