Hacker NewsWhistler: Live eBPF Programming from the Common Lisp REPL
https://atgreen.github.io/repl-yell/posts/whistler/
#HackerNews #Whistler #eBPF #CommonLisp #REPL #Programming #LiveCoding
GripNews๐ GitHub - gcv/julia-snail๏ผEmacs ็ Julia ้็ผ็ฐๅข
โค ็บ Emacs ไฝฟ็จ่ ๆ้ ็้ซๆ Julia ไบคไบๅผ้็ผ็ฐๅข
โค https://github.com/gcv/julia-snail
Julia Snail ๆฏไธๅฅๅฐ็บ Emacs ่จญ่จ็ Julia ้็ผ่ REPL ไบๅๅฅไปถใๅ ถ้็ผ็ๅฟต่ดๆฌไบ Common Lisp ็ SLIME ่ Clojure ็ CIDER๏ผๆจๅจ็บ Julia ๆไพ้ซๆใ้ๆดป็ไบคไบๅผ้็ผ้ซ้ฉใSnail ๅฉ็จ้ซๆ่ฝ็ต็ซฏๆจกๆฌๅจ๏ผvterm ๆ Eat๏ผไพๅ็พๅ็ REPL๏ผๆๆ่งฃๆฑบไบๅณ็ตฑ Emacs ็ทฉ่กๅ้กฏ็คบ็ฐๅธธ็ๅ้กใ้้่ Emacs ๅ็ xref ็ณป็ตฑใ่ชๅ่ฃๅ จๆฉๅถๅ CSTParser ็ๆทฑๅบฆๆดๅ๏ผ้็ผ่ ่ฝๆด็ด่งๅฐๅท่ก็จๅผ็ขผ่ผๅ ฅใๅฎ็พฉ่ทณ่ฝๅ็ฌฆ่่ฃๅ จ๏ผๅฏฆ็พ็ก็ธซ็้็ผๅทฅไฝๆตใ
+ ็ตๆผๆๅ CIDER ้ฃๆจฃ็ๅทฅๅ ทไบ๏ผๅฐๆผ็ฟๆ ฃๅจ Emacs ่ฃกๅฏซ Clojure ็ไบบไพ่ชช๏ผJulia Snail ่ฎๅๆ่ช่จ่ฎๅพ่ผ้ฌไธๅฐใ
+ ้็ถๅฎ่ฃ v
#Emacs #Julia #่ป้ซ้็ผ #้ๆบๅทฅๅ ท #REPL
โค ็บ Emacs ไฝฟ็จ่ ๆ้ ็้ซๆ Julia ไบคไบๅผ้็ผ็ฐๅข
โค https://github.com/gcv/julia-snail
Julia Snail ๆฏไธๅฅๅฐ็บ Emacs ่จญ่จ็ Julia ้็ผ่ REPL ไบๅๅฅไปถใๅ ถ้็ผ็ๅฟต่ดๆฌไบ Common Lisp ็ SLIME ่ Clojure ็ CIDER๏ผๆจๅจ็บ Julia ๆไพ้ซๆใ้ๆดป็ไบคไบๅผ้็ผ้ซ้ฉใSnail ๅฉ็จ้ซๆ่ฝ็ต็ซฏๆจกๆฌๅจ๏ผvterm ๆ Eat๏ผไพๅ็พๅ็ REPL๏ผๆๆ่งฃๆฑบไบๅณ็ตฑ Emacs ็ทฉ่กๅ้กฏ็คบ็ฐๅธธ็ๅ้กใ้้่ Emacs ๅ็ xref ็ณป็ตฑใ่ชๅ่ฃๅ จๆฉๅถๅ CSTParser ็ๆทฑๅบฆๆดๅ๏ผ้็ผ่ ่ฝๆด็ด่งๅฐๅท่ก็จๅผ็ขผ่ผๅ ฅใๅฎ็พฉ่ทณ่ฝๅ็ฌฆ่่ฃๅ จ๏ผๅฏฆ็พ็ก็ธซ็้็ผๅทฅไฝๆตใ
+ ็ตๆผๆๅ CIDER ้ฃๆจฃ็ๅทฅๅ ทไบ๏ผๅฐๆผ็ฟๆ ฃๅจ Emacs ่ฃกๅฏซ Clojure ็ไบบไพ่ชช๏ผJulia Snail ่ฎๅๆ่ช่จ่ฎๅพ่ผ้ฌไธๅฐใ
+ ้็ถๅฎ่ฃ v
#Emacs #Julia #่ป้ซ้็ผ #้ๆบๅทฅๅ ท #REPL
screwlisphttps://lispy-gopher-show.itch.io/leonardo-calculus/devlog/1451887/my-ansi-common-lisp-condition-ontology-eg
#devlog (!) #commonLisp #programming #article on my #itchio .
Since I forgot I had to ping aral to let my blog webhook work again, here is this article exhibiting
The ANSI common lisp condition system
in particular, I wrote an #ontology for classical! #lisp expression generation supporting precondition .. postcondition = handler .. local restart
interactive #repl exploration. The blog will resume its webhook updates next week. I should love this itch already anyway.
My ANSI common lisp condition ontology eg - Leonardo Calculus Software Individuals by screwtape
Fallow week oh fallow week. I mean, the show with cdg, kmp, and Ramin Honary then Kent Pitmanโs epic ensuing mastodon thread featuring Pitman, Ramin, Gosling, Doug Merritt, and Roger Crew and others...
ArneBab@cwebber So to fix that, let me tell you about the PR for spritely hoot-repl that reduces load times of the #Guile #Scheme web #REPL in #webassembly by at least 30% โบ
https://codeberg.org/spritely/hoot-repl/pulls/4
Though Iโm sure you already know, so this is just an "I answered the review" notification, but more interesting than something about LLM agents โบ
Optimize load time
This change parallelizes all resource loading. Before this change, all resources were loaded by the browser sequentially:First reflect.js and repl.js. After those are finished, wtf8.wasm (triggered from repl.js, initiated from reflect.js). Then repl.wasm (same). And finally (after a few ms pro...
pancsta@lobsters Lets not forget about debugging - currently it's either print log (crap) or a full trace (heavyweight). Im addressing this in #golang with a #TUI debugger with time travel and a #REPL https://asyncmachine.dev/wasm but that still doesnt replace actual breakpoints
The new release of #asyncmachine brings #WASM support and browser compat - including #aRPC, #TUI debugging and #REPL. Check out the changelog for v0.18.0 https://github.com/pancsta/asyncmachine-go/releases/tag/v0.18.0 and a dedicated WASM example https://asyncmachine.dev/wasm (with D2 diagrams).
This allows to have a single #statemachine #distributed across n servers and m browsers, using efficient diff-based synchronization.
Looking fwd to #wasmio!
sayzardAwni Hannun (@awnihannun)
mlx_lm.server๋ฅผ ์ฌ์ฉํด ๋ก์ปฌ์์ ์คํ ๊ฐ๋ฅํ LM(REPL ํฌํจ) ์์ ๋ฅผ ๊ณต๊ฐํ๋ค. Qwen3 Coder Next ๋ชจ๋ธ์ M3 Ultra์์ ๊ตฌ๋ํ๋ฉฐ, ๋ชจ๋ธ์ด ์ฐ์ฐ์ ์์ฒญํ๋ฉด ์ฝ๋ ์์ฑ โ ์คํ โ ๊ฒฐ๊ณผ ๋ฐํ ๊ณผ์ ์ ๋ฐ๋ณตํ๋ ๊ตฌ์กฐ๋ฅผ ๋ณด์ฌ์คฌ๋ค. ์ด๋ ๋ชจ๋ธ์ ๋ก์ปฌ ์คํ ๋ฐ ์ฝ๋ ์์ฑ ํ ์คํธ์ ์ ์ฉํ ๊ฐ๋จํ ์ฌ๋ก๋ก, ๊ฐ๋ฐ์๋ค์ด ์คํ๋ผ์ธ ํ๊ฒฝ์์๋ ์คํํ ์ ์๋๋ก ๋๋๋ค.
Awni Hannun (@awnihannun) on X
Made a simple example of an LM with a REPL using mlx_lm.server to run locally (using Qwen3 Coder Next on an M3 Ultra). 1. Ask model to compute something (here 1000th fibonacci number). 2. Model generates the code 3. Run the code in the REPL 4. Get result and return to step 2 if
Awni Hannun (@awnihannun)
'Recursive LM' ๋ ผ๋ฌธ ๊ด๋ จ ์ถ๊ฐ ๋ ผ์์์ ํต์ฌ์ ํ๋กฌํํธ๋ฅผ ์ธ๋ถํํด ํ์ LLM๋ค์ด ๊ฐ๊ฐ ๋ถ๋ถ ์์ ์ ์ํํ๊ณ ๊ฒฐ๊ณผ๋ฅผ ํตํฉํ๋ ์ฌ๊ท์ ๊ตฌ์กฐ๋ผ๋ ์ ์ด ๊ฐ์กฐ๋ฉ๋๋ค. ๋ํ LLM์ REPL์ ์ ๊ณตํ์ฌ ์ฝ๋ ์คํ๊ณผ ๊ฐ์ ์ํธ์์ฉ์ ๊ฐ๋ฅํ๊ฒ ํ๋ ๋ถ๋ถ์ด ์ฃผ์ํ ํ์ ํฌ์ธํธ๋ก ์ธ๊ธ๋ฉ๋๋ค.
Awni Hannun (@awnihannun) on X
Some replies along the lines of the key idea is breaking down the prompt and recursively running sub LLMs on it and stitching them back together. I understand thatโs the central premise of the paper. But the lasting nugget is giving the LLM a REPL (which may not even be novel in
Awni Hannun (@awnihannun)
์ ๋ ผ๋ฌธ 'Recursive LM (RLM)'์ด ํ์ ์ ๋๋ค. ํต์ฌ ์์ด๋์ด๋ ์ธ์ด๋ชจ๋ธ(LM)์ REPL ํ๊ฒฝ์ ์ ๊ณตํ๋ค๋ ์ ์ผ๋ก, ์ด๋ฅผ ํตํด ๋ชจ๋ธ์ด ์ฝ๋๋ฅผ ์คํํ๋ฉด์ ๋ฌธ์ ๋ฅผ ๋จ๊ณ์ ์ผ๋ก ํด๊ฒฐํ ์ ์์ต๋๋ค. ๋ ผ๋ฌธ์ ์ฅ๋ฌธ ๋ฌธ๋งฅ ์ฒ๋ฆฌ ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ๋ค๊ณ ์๋ ค์ก์ผ๋, ์ค์ ๋ก๋ LM๊ณผ REPL์ ๊ฒฐํฉ์ด ๊ฐ์ฅ ํฅ๋ฏธ๋ก์ด ํ์ ์ผ๋ก ํ๊ฐ๋ฉ๋๋ค.
Awni Hannun (@awnihannun) on X
Looking at the recursive LM (RLM) paper this morning. It's actually quite simple and nice idea: give the LM a REPL. The paper is marketed as solving long-context. But I think the key nugget is to give the LM a REPL. The REPL is useful because: - Execute code in it -> let's you
Awni Hannun (@awnihannun)
mlx_lm.server๋ฅผ ์ฌ์ฉํด ๋ก์ปฌ์์ ์คํ๋๋ ์ธ์ด ๋ชจ๋ธ(REPL ๊ธฐ๋ฐ) ์์๋ฅผ ๊ณต์ ํ์ต๋๋ค. Qwen3 Coder Next ๋ชจ๋ธ์ M3 Ultra ์นฉ์์ ๊ตฌ๋ํ๋ฉฐ, ๋ชจ๋ธ์ด ์ฝ๋๋ฅผ ์์ฑํ๊ณ REPL์์ ์คํํ ํ ๊ฒฐ๊ณผ๋ฅผ ๋ฐํํ๋ ๊ณผ์ ์ ์์ฐํ์ต๋๋ค. ์ด๋ ๋ก์ปฌ ํ๊ฒฝ์์ LM ๊ธฐ๋ฅ์ ์คํํ๊ณ ์ ํ๋ ๊ฐ๋ฐ์๋ค์๊ฒ ์ ์ฉํ ์์๋ก ๋ณด์ ๋๋ค.
Awni Hannun (@awnihannun) on X
Made a simple example of an LM with a REPL using mlx_lm.server to run locally (using Qwen3 Coder Next on an M3 Ultra). 1. Ask model to compute something (here 1000th fibonacci number). 2. Model generates the code 3. Run the code in the REPL 4. Get result and return to step 2 if