Ian Sweet

Check out our ✨ fantastic ✨ list of accepted papers (pldi25.sigplan.org/track/pldi-202…)! You still have time to book your tickets to Seoul and watch the PLDI talks live!

We are hosting the MIT Programming Languages Review on April 25th in person here at MIT! The PLR is a student-run workshop that aims to highlight the best papers from the past year that we believe will have a significant impact on shaping the future direction of PL research.

@mitchellh @metafrost @sharno3 @spiralladder_ Anyway, appreciate and admire your work and I hope that your experience improves! It totally sucks waiting for something to compile, especially when you want to iterate quickly and rely on the compiler to help you.

@mitchellh @metafrost @sharno3 @spiralladder_ By analogy, C compiles can be very slow with poorly crafted Makefiles that include stuff over and over. That’s a joint blame — the developer could have written a better Makefile, and/or the C compiler could be smarter during preprocessing.

Is the Rust project doing anything about the insanely slow compile times? I see various blurbs about language features, build tooling, lib enhancements, but I rarely see the same activity around compiler speed. Genuine question from an outsider, these compiler speeds are killing me as a consumer, I can't imagine what it's like as a developer.

@VictorTaelin I see, DeepSeq is just the decider. I think my original two questions still stand, though.

@VictorTaelin Finally, why can’t you parallelize in the same way to deal with the issue of context size? Just chunk the codebase in such a way that chunk size fits into context. No?

IT WORKS! Demo time 🥳 (next 10x productivity bump is here?) Suppose you must refactor a large codebase, e.g.: > "use I32 instead of U32 as the native number type" That task, by itself, is simple enough to be automated by Sonnet alone, right? Problem is: what if the codebase is too large to fit on the AI's context, and RAG solutions / editors like Cursor can't precisely isolate which parts require edit (because it needs some semantic reasoning)? What now? Do we fallback to human coding, as did the old Mayans? Of course not! We just split the codebase in chunks, and send each chunk to ~500 DeepSeek instances in parallel (!). Each instance then decides if given block should be edited or not. This is viable because DeepSeek is smart, fast and really cheap. Finally, we send the blocks that need edit to Sonnet, in a single prompt, and let it work normally. The entire process takes less than a minute, is cheap enough, and very effective. We conclude with a 'git diff' for manual review. In short, AI has came to grep search & replace. Since this script feels like dropping a bomb in your codebase, I'm calling it AoE, and it is available on VictorTaelin/AI-Scripts. Enjoy! Here's the demo:

My tierlist for the PoE2 ascendancies. First pic: MINIMUM tier expected for an ascendancy (pessimistic evaluation, given that we don't know a lot) 2nd pic: likelyhood for a tier to be better than tiered. However: that's a risk/chance. Do you agree? What's your S tiers?

A fun page on some of the kinds of performance engineering work we do at Jane Street. janestreet.com/performance-en…

@josecalderon For real, it is integral to so many engineering disciplines!

This is a very surprising opinion for a guy with continuously bad takes.

@notypes - Abstract Interpretation - Module Systems - Contextual Equivalence && (Stuttering) Refinement

@djg98115 +1 for Matt Might’s blog posts on the topic. For folks who are familiar with SSA they could read: * cs.princeton.edu/~appel/papers/… * cs.purdue.edu/homes/suresh/5… Then, follow-up with all the literature on ANF, CPS, e.g. Andrew Kennedy’s work: microsoft.com/en-us/research…

Are seminal papers approachable in 2024? Prefer mechanics and pragmatics over denotations and metatheory? Are there primers or textbook chapters? Imagine a few hours of reading (1-2 papers or chapters) not a whole course (as awesome as that would be). TIA (2/2)

PL-research literature question: Imagine a 2024 grad student familiar with operational semantics, types, structural induction but had never heard of continuation-passing style or cps-transform. Need to learn them and the key properties of cps. What should they read? (1/2)

@ChShersh Source? I’d be surprised if entropy was an issue. Password generation happens infrequently enough that even very slow entropy generation ought to keep up… no?

Security tip for free: Never share which password manager you use. Some of them have shitty defaults (like not enough entropy), and a dedicated person can crack your account easily. Not to mention known vulnerabilities. Generally, the less you share about your security solution, the better.

@rg9119 Ah yes, the classic one-two punch of survivorship bias and stockholm syndrome.

"since we understand the meaning of our notation sufficiently well." :: commences endless screaming :: (from Knuth: BIG OMICRON AND BIG OMEGA AND BIG THETA)

@yminsky I won’t be at ICFP but I’m getting serious shirt FOMO. This is fantastic.

Our new t-shirt. We're very excited for multicore OCaml...

@jeanqasaur AGI doomers: …but is it mindful?

The problem with a lot of AI-generated art is that it's neither cutesy nor demure

@andrewkhirsch Have some 10k foot view highlights? One thing that comes to mind is that macros make implementing static EPP easy. Anything else? Any nifty uses of continuations?

I'm happy to announce that my student Alex Bohosian just released Choret, our implementation of choreographies in Racket! Please take a look, and feel free to ask any questions/provide any feedback here: github.com/Foundations-of…