Yoav

1.6K posts

Yoav banner
Yoav

Yoav

@YoavCodes

Technical founder. Building Electrobun and Dash at https://t.co/Pp6sheAD3q | 20+ years in early startups going zero to one | prev: 2015-2022 @webflow

Katılım Mayıs 2025
77 Takip Edilen1.5K Takipçiler
Yoav
Yoav@YoavCodes·
Imagine deploying 1,000,000 lines of code written in 6 days by AI that no human has ever read, let alone reviewed, to production where your customer’s data is. Imagine
Ryan Fleury@rfleury

Presented without comment

English
89
120
3.5K
289.5K
Yoav
Yoav@YoavCodes·
I’m hoping it’s just a miscalculation from being inside the current SF bubble. I’ve already had to start working on removing bun from electrobun by making it optional and not core. It’s really unfortunate. I’m just asking for humans to look at the code until it’s proven they don’t have to.
English
0
1
6
239
João Mendonça
João Mendonça@joaomendoncaaaa·
@YoavCodes @jarredsumner @Will_I_AmLegend brother, as the biggest LLM industrial complex hater, do you really think they don't care...? it's a matter of incentives + tradeoffs this was the path they chose, the path doesn't end here
English
1
0
1
129
perfect gentleman
perfect gentleman@Will_I_AmLegend·
Still waiting for that viral post where someone shows the new vibe-coded, AI slop, lazy rewrite of bun actually failing at any task... @jarredsumner
English
3
2
52
9.4K
Yoav
Yoav@YoavCodes·
In a different tweet you mentioned the Overton window. If you have access to advanced models that don’t hallucinate as often as the ones we all have access to it would have cost your team nothing to prove that by reading the code and letting it marinate for longer until people catch up. But model cards and experience say it’s unlikely. Choosing to roll it out this way means people can’t even pin to the last zig version while it marinates because any zero day discovered would force any company to choose between getting a security patch at the expense of undefined behaviour. Any rational person has to stop using Bun in production now. It doesn’t have to be this way. It’s not too late to take a more rational path.
English
0
2
8
301
Yoav
Yoav@YoavCodes·
The concern isn’t that it will break things, the unit tests pass. The concern is how many hallucinations or misaligned additional code was added. What was previously believed to be a battle tested thing that’s had lots of eyes on it running on my server handling my customer’s data. Is now a black box that presumably no one has really looked at, that’s going to maintain a velocity and practice of no one looking at. What’s in the blackbox handling my customer’s data? How can Bun continue to be core infrastructure going forward? I wish it was at least kept as canary for long enough for the community to plausibly look at what was generated. Instead of the next version of bun being this new unseen thing.
English
2
2
12
986
Yoav
Yoav@YoavCodes·
@raztomouse @syrusakbary They said the last zig version has already shipped, the next version will be rust, and they’re just using more ai to review the new million lines. It’ll be a black box that happens to pass the unit tests and do who knows what else.
English
0
0
0
28
Cåt
Cåt@raztomouse·
@YoavCodes @syrusakbary Woah, has bun made a release based on the rust fork already? That's crazy
English
1
0
0
32
Syrus Akbary — eu/acc
Syrus Akbary — eu/acc@syrusakbary·
Soon, you will be able to run Node.js fully using Quick.js as the JavaScript Engine (instead of V8). Stay tuned!
English
1
0
8
492
Yoav retweetledi
Miguel Salinas
Miguel Salinas@Vercantez·
Coming off of this psychosis and it’s painful. Not shipping any more features until I can explain every part our codebase in detail. Especially important to do before we open source
Mitchell Hashimoto@mitchellh

I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they include personal friends I deeply respect, but I worry about how this plays out. I lived through the great MTBF vs MTTR (mean-time-between-failure vs. mean-time-to-recovery) reckoning of infrastructure during the transition to cloud and cloud automation. All those arguments are rearing their ugly heads again but now its... the whole software development industry (maybe the whole world, really). It's frightening, because the psychosis folks operate under an almost absolute "MTTR is all you need" mentality: "its fine to ship bugs because the agents will fix them so quickly and at a scale humans can't do!" We learned in infrastructure that MTTR is great but you can't yeet resilient systems entirely. The main issue is I don't even know how to bring this up to people I know personally, because bringing this topic up leads to immediately dismissals like "no no, it has full test coverage" or "bug reports are going down" or something, which just don't paint the whole picture. We already learned this lesson once in infrastructure: you can automate yourself into a very resilient catastrophe machine. Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying. I worry.

English
12
19
730
60K
Yoav
Yoav@YoavCodes·
@scubzero @ee0pdt It’s definitely not ok for all the people running bun in production that now have to adjust roadmaps to deal with this. But I agree they can do whatever they want and should, and the ecosystem will just replace it and move on.
English
1
0
1
37
Dev No-ops
Dev No-ops@scubzero·
@ee0pdt @YoavCodes I think this is ok, but to me if someone introduced a brand new JavaScript runtime would have to be in production for the bleeding edge users for at least 6 months before I would consider it for prod workloads. This is essentially a brand new runtime to me.
English
1
0
0
34
Yoav
Yoav@YoavCodes·
If Claude was so good, they could have had a rust port as a separate repo, and had claude port every PR. Or just have it fix the segfaults it’s been introducing the last few months since acquisition. There is no reason to YOLO +1M rust, -700k zig as a minor version bump in 6 days and announce no human will ever look at the code That just kills bun for any serious person.
Yoav@YoavCodes

Imagine deploying 1,000,000 lines of code written in 6 days by AI that no human has ever read, let alone reviewed, to production where your customer’s data is. Imagine

English
11
25
567
39.8K
Yoav
Yoav@YoavCodes·
@jayair This will be remembered the way we remember Theranos.
English
0
0
9
1.8K
Jay
Jay@jayair·
Bun is probably the most intense piece of software created in the last five years, where its moat was its difficulty Rewriting it is so crazy that you wouldn't even consider it I think this'll be remembered as one of the defining moments in the history of coding models
English
30
7
398
34.9K
Yoav retweetledi
Mitchell Hashimoto
Mitchell Hashimoto@mitchellh·
I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they include personal friends I deeply respect, but I worry about how this plays out. I lived through the great MTBF vs MTTR (mean-time-between-failure vs. mean-time-to-recovery) reckoning of infrastructure during the transition to cloud and cloud automation. All those arguments are rearing their ugly heads again but now its... the whole software development industry (maybe the whole world, really). It's frightening, because the psychosis folks operate under an almost absolute "MTTR is all you need" mentality: "its fine to ship bugs because the agents will fix them so quickly and at a scale humans can't do!" We learned in infrastructure that MTTR is great but you can't yeet resilient systems entirely. The main issue is I don't even know how to bring this up to people I know personally, because bringing this topic up leads to immediately dismissals like "no no, it has full test coverage" or "bug reports are going down" or something, which just don't paint the whole picture. We already learned this lesson once in infrastructure: you can automate yourself into a very resilient catastrophe machine. Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying. I worry.
English
401
1.4K
11.7K
958.6K
Yoav
Yoav@YoavCodes·
@Jack1000k Everyone who currently runs bun on their servers, and has customers.
English
0
0
0
11
Dingbat472
Dingbat472@dingbat472·
@YoavCodes chances are 50% of these edits arent actual code edits, but claude re-writing the code 1 line down (with line-endings not set to be accounted for and ignored), causing a shift in all the code thus a billion diffs. Had this happen with one of my repos as i swapped from mac to win
English
1
0
0
608
Yoav
Yoav@YoavCodes·
@tomschrupp Except to an LLM comments are code, as they keep vibe coding this new monstrosity.
English
0
0
1
18
Tom Schrupp
Tom Schrupp@tomschrupp·
@YoavCodes To be fair, if my recent experience with Claude is any indication, 90% of that "code" is just bloated comments
English
1
0
1
59