Bun posts Rust porting guide, says rewrite is still half-baked

Bun creator Jarred Sumner has posted a Zig-to-Rust porting guide, igniting speculation that the project may migrate away from Zig, though Sumner said there is no commitment to rewriting, only that he is "curious to see what a working version of this looks like." Bun, a JavaScript runtime and toolkit, is a prominent user of Zig, …

[Anonymous Coward]

Seeing a project written in Zig or Go always gave me the mental image of someone entering a room, ripping a loud fart and leaving.

[lotus123]

>"someone entering a room, ripping a loud fart and leaving"

Actually seeing post like this involves exactly the same impression

[jake]

I assume that by "like this" you mean your own post?

[CRConrad]

I think you assume wrong.

I'm fairly sure they meant “like the original AC’s comment”.

[QET]

"

Whether Bun migrates to Rust or not, Sumner is convinced that open source software (OSS) will make increasing use of AI. Commenting on Zig's AI ban, he said on X: "I expect OSS to go the opposite direction: no human contribution allowed." People will still discuss issues and prioritization, he said, "but the actual act of writing code, of submitting PRs, of replaying and addressing feedback, of implementation will be LLMs."

"

Is he completely ignoring the fact many flock to OSS due to the high amount of focus on ethics?

Ethics like "AI" trained on content without explicit and verbose permission for each and every bit of code-base hoovered up for training is not only prohibited, but hard-stop one of the biggest sins imaginable?

[QET]

With a cherrypicking flavored whataboutism comment like that, I can safely say I see why you chose to be AC...

[FIA]

Ethics like "AI" trained on content without explicit and verbose permission for each and every bit of code-base hoovered up for training is not only prohibited, but hard-stop one of the biggest sins imaginable?

If that's is the biggest sin you can imagine then your life is good and your imagination limited.

[Anonymous Coward]

Long term, it's just not viable to maintain a strict, zero-tolerance AI policy.

Currently, developers have good reason to be much more skeptical of AI-generated code. It does cause problems. The time saved by the submitter gets lost in code review. Bugs and hallucinations often fly under the radar.

Yet, some AI-assisted submissions are worthwhile. Worthwhile enough? TBD. Much like early, bug-ridden computers, AI is improving. Will it get better by enough, and when? TBD.

Even if they're not comfortable with AI right now, the smart thing to do is begin the long conversation about what must be true before AI can be used. That's not an easy subject and it won't happen overnight.

No different than discussing which safety benchmarks must be met by self-driving cars before they're allowed to operate fully autonomously. Clear benchmarks set clear development goals.

A straight ban is more likely to be entirely repealed, or replaced with rushed and weak safety criteria, thus resulting in less safe outcomes than if safety criteria are set well in advance of any specific vendor's deployment.

[jake]

"no human contribution allowed"

One wonders what flavo(u)r the Cool-aid was ...

[Bebu sa Ware]

Re: "no human contribution allowed"

One wonders what flavo(u)r the Cool-aid was ...

Definitely the flavour that needs to be swallowed with a sizeable dose of the bursar's dried frog pills.

[QET]

Re: "no human contribution allowed"

That image is how I imagine someone convinced Elon Musk to do all the drugs that he does.

[Anonymous Coward]

Re: "no human contribution allowed"

Golden. This is a smart statement if you want someone to buy you out as a billion dollar project. Don't hate the player, hate ...

[jake]

Re: "no human contribution allowed"

Golden? Not a lot of liquids are golden in colo(u)r ... although I can think of one.

_{Before buying into the scam, ask yourself why yer DOM advised you not to eat yellow snow.}

[DrewPH]

> Sumner is convinced that open source software (OSS) will make increasing use of AI. Commenting on Zig's AI ban, he said on X: "I expect OSS to go the opposite direction: no human contribution allowed."

He's clearly hallucinating as much as the LLM models he loves so much.

[elsergiovolador]

No calculator

No calculator maths please.

[CRConrad]

On the contrary: More calculator maths, please.

I'm still baffled as to why “AIs” — actually, LLMs — still refuse to ”use a calculator”, i.e. call a pre-compiled arithmetics subroutine whenever they encounter an arithmetics problem, but insist on doing the maths “in their head”; their hallucinating “head”.

Like, apparently they're able to recognise that something _is_ arithmetic, since they know to paste in an arithmetic-sounding answer. And they're a program running on a computer, sometimes able to call external programs, and in any case able to have a deterministic callable arithmetics subroutine built-in by their creators. Any science fiction AI anyone has ever heard of, from Star Trek (at least TNG) to The Culture and beyond, does it that way.

So how moronic are the people who build these things — and so often profess to be huge SF fans since childhood — not to build them that way? Sure seems we aren't going to see “AGI” any time soon: As long as these things don't know when to reach for a “calculator”, the “I” will continue to stand not for “Intelligence”, but rather its opposite.

[RLWatkins]

"Most open-source code will be written by AI"

No, no it won't. We require that it work properly, so that option is off the table.

[Pete 2]

Deal breaker

> make breaking changes to the language

And I would suggest that most serious organisations will therefore steer clear. Until either the rate of breakage drops to zero or someone has a quiet word with the guy. To explain what backwards compatibility means, and why it is the secret to commercial success. Assuming commercial success is even on the radar.