Jarred Sumner

26.1K posts

Jarred Sumner

Jarred Sumner

@jarredsumner

building @bunjavascript at @anthropicai. formerly: @stripe (twice) @thielfellowship. high school dropout. npm i -g bun

San Francisco, CA Katılım Mayıs 2014
643 Takip Edilen175.1K Takipçiler
mattn
mattn@mattn_jp·
bun の stringWidth、Rust のコードが消されて C++ のコードが追加される。 やっぱり C++ が勝つんだね。 github.com/oven-sh/bun/pu…
日本語
12
17
251
46.8K
Nick
Nick@nickcammarata·
if you want a 2x speed up on something and you ask an agent it’ll come back with sorry the best I could find was 1.5x :( if you instead ask for 50x it’ll come back saying sorry the best I could find was 30x :( then use that one
English
19
13
709
53.8K
Jarred Sumner
Jarred Sumner@jarredsumner·
@NyxXeres @vegai_ @demfabris > it wasn’t vetted, if any effort was put into the review it was … This is false. I spent many hours/days watching & reading & monitoring the code live, and coordinating large refactors. This idea - that I just ran the test suite and merged is completely wrong.
English
1
0
6
276
Nyx
Nyx@NyxXeres·
@vegai_ @demfabris @jarredsumner The issue isnt really it was generated with AI, the issue is that it wasn’t vetted, if any effort was put into the review it was limited to “hey, the test suite passes, I guess it’s 100% the same logic, I did tell the LLM make no mistakes afterall”
English
1
0
0
221
Jarred Sumner
Jarred Sumner@jarredsumner·
@ShomLinEd More accurate to say claude made it tbh. It’s making it coverage-guided now.
English
0
0
9
719
Shom
Shom@ShomLinEd·
@jarredsumner Is this fuzzer a library or made by you?
English
1
0
0
682
Jarred Sumner
Jarred Sumner@jarredsumner·
@zeeg the game is to set up the agent for success (usually using agents) first. it needs to be able to end to end build & test & verify the code works. if it can’t see logs, if it can’t run the app, if it can’t see errors, if it can’t write tests, it’s not going to work well.
English
2
1
77
4.9K
David Cramer
David Cramer@zeeg·
This is only true in the most specific of scenarios. It is very hard to test many types of systems, and in those cases the LLMs are degrading the reliability of projects. There’s certainly a world where you dive in and do a bunch of work manually to get it into a good spot, but in my experience (and many of my peers) they are very very poor at managing software quality. You especially see this in things like agents themselves, or scenarios that require a lot of complex network interactions. It’s why I keep investing into evals-as-tests, but it’s still not been reliable enough. I’m sure it’ll get better but let’s be realistic about the challenges.
Uncle Bob Martin@unclebobmartin

I am absolutely more productive using agents. I don't know the factor but it's large. However much of that productivity is spent tuning the agents and hardening the product. I'm guessing 30%-40%. Some might consider that a waste; but I don't. The software I'm creating nowadays is vastly more robust than I'd ever been able to create manually. I don't mean that the code is better. I mean the surrounding tests are vastly better. I have a higher degree of confidence than I ever had manually -- even when I used very disciplined TDD and Acceptance testing. And then there's the ability to quickly reorganize the modules and the architecture while keeping those robust tests running. That is a tremendous boon.

English
10
2
86
22.3K
Jarred Sumner
Jarred Sumner@jarredsumner·
I’m very motivated to prove the people making “jokes” about the Rust rewrite introducing security vulnerabilities wrong.
English
10
12
279
38.5K
Jarred Sumner
Jarred Sumner@jarredsumner·
For the JS parser fuzzer, I set it up to use swc’s test suite as the corpus and then change & permute “interesting” characters. Their existing tests already cover lots of edgecases, so it’s an especially good one to fuzz with.
English
2
0
126
10.8K
Jarred Sumner
Jarred Sumner@jarredsumner·
@gabriel1 yeah some years ago I switched to only listening to music while running and that was a good productivity boost
English
0
0
21
2K
@levelsio
@levelsio@levelsio·
I know Anthropic has a GPU shortage but every day forcefully putting my effort back to medium feels...well....annoying
@levelsio tweet media
English
163
20
1.3K
117.3K
Jarred Sumner
Jarred Sumner@jarredsumner·
@Malix_Labs Not at all. The baseline was former Bun. The rewrite changed the language & changed very little else.
English
4
0
109
7.7K
Jarred Sumner
Jarred Sumner@jarredsumner·
@theo In other words, it’s more accurate to say I ported Bun to Rust for the compiler-assisted tooling to help us fix bugs than I moved away from Zig. It was never anti-Zig. Programming languages are now a mutable technical decision you can change in like a week.
English
3
4
209
10.9K
Jarred Sumner
Jarred Sumner@jarredsumner·
@theo I agree Who knows maybe Bun will be in some yet-to-be-written language next year. Or maybe Bun will simultaneously be in Java, in Cobol, in Ruby, and Python all at the same time, for fun. Like doom or bad apple
English
10
0
414
25.4K
Theo - t3.gg
Theo - t3.gg@theo·
A lot of people are building with the assumption that the codebases we work in today will still matter next year. I’m not sure if that’s the case.
English
64
65
2.3K
218.1K
Jarred Sumner
Jarred Sumner@jarredsumner·
@50onice > don’t share publicly This is not accurate. It’s always been a public repo. It is linked in our docs. Why do people make stuff up like this github.com/oven-sh/webkit
English
7
5
1.1K
35.1K
Barrett Harber
Barrett Harber@50onice·
Bun has a private build of WebKit that’s optimized for size and they don’t share it publicly, but it’s in the release binary
English
5
0
61
32.3K
Jarred Sumner
Jarred Sumner@jarredsumner·
@meekaale @theo Today there’s no real reason to. JSC doesn’t have the same scale of issues we were having. If all this gets cheap enough to not care at all then maybe but it wouldn’t make sense to lose their upstream progress
English
0
0
19
1.8K
Arav
Arav@AravPhi·
@jarredsumner @rustaceans_rs cmon guys its defiantly harder to write a blog post about porting bun to rust than actually porting bun to rust
English
3
1
79
14.4K
Rust Bytes 🦀
Rust Bytes 🦀@rustaceans_rs·
what are you working on this weekend?
English
19
1
102
25.9K
Jarred Sumner
Jarred Sumner@jarredsumner·
@DDrejer We fixed numerous memory leaks in worker_threads / Worker, the bundler, and hot reloading at runtime. These are noticeable. There were some tiny memory leaks in rare error-handling code paths across the runtime. Those probably do add up if running without redeploying for months
English
1
0
23
4.7K
Jarred Sumner
Jarred Sumner@jarredsumner·
Bun in Rust is better than the original, and it’s going to keep getting better. We fixed a lot of bugs inherited from the original. We’re fuzzing a lot more.
Bun@bunjavascript

🤏 update:

English
74
54
1.9K
513.2K
Jarred Sumner
Jarred Sumner@jarredsumner·
@notnullptr A lot of automated crash reports on Windows have anti-cheat DLLs in the stack trace. Very hard for us to do anything about it. Chrome has a hardcoded list of these where they basically ignore crash reports from.
English
5
1
316
20.3K
nullptr 🐱🍩
nullptr 🐱🍩@notnullptr·
In the next version of Bun, we're introducing a full Vanguard bypass
English
4
8
333
30K
Jarred Sumner
Jarred Sumner@jarredsumner·
@paulbohm @50onice Relocations make the binary larger, so it’s unlikely we’ll do this. Cool project tho
English
1
0
20
3K

Keşfet

@mattn_jp @nickcammarata @NyxXeres @vegai_ @demfabris @ShomLinEd @zeeg @gabriel1