4 ms·

Or he's just giving a sane take, one that most people in the Bay Area have by now. Lots of tech companies are doing just fine with purely AI written code at th

Or he's just giving a sane take, one that most people in the Bay Area have by now. Lots of tech companies are doing just fine with purely AI written code at this point. Saying that the quality is getting worse in some immeasurable way (while incidents remain the same) is literally unfalsifiable.

vnaghra 6mo ago
It is getting worse in a measurable way: 1) Since vibe coding GitHub has frequent outages and isn't able to load a large number of comments. 2) The slop translation of Bun resulted in immediate bugs (https://github.com/oven-sh/bun/issues/30719 https://github.com/oven-sh/bun/issues/30719) that the hyped Mythos apparently did not find. 3) AI features and (likely, though not proven) AI code resulted in a 0-day in Google code: https://projectzero.google/2026/01/pixel-0-click-part-1.html https://projectzero.google/2026/01/pixel-0-click-part-1.html The house of cards is beginning to collapse.
- sebzim4500 6mo ago
  Maybe this is confirmation bias but I feel like Github has had terrible uptime since before LLMs were a thing.
- foldr 6mo ago
  People started depending on GitHub more. Do people really think it was more reliable when it was a sprawling RoR app in 2010? (Not that there’s anything wrong with RoR; people just didn’t expect such high uptime back then.)
- vividfrier 6mo ago
  [flagged]
armchairhacker 6mo ago
> Lots of tech companies are doing just fine with purely AI written code at this point. If there are so many, surely you or one of the other AI supporters have a public example? I’m aware of two examples, although they’re (mostly routine) translation with existing test infrastructure, so easier for an LLM: - Bun’s rewrite, although we haven’t seen the effects on further development - Ladybird’s rewrite, which seems to be continuing fine
3form 6mo ago
>Or he's just giving a sane take, one that most people in the Bay Area have by now. I don't know what the Bay Area note is supposed to mean in the context of the whole post - unless you want to reinforce that it surely means that it's a sane take... In which case, I'm not certain the non-Bay readers would agree that it comes from an unbiased culture.
arkadiytehgraet 6mo ago
This account is another LLM-hype peddler, shilling specifically for Anthropic.
- solenoid0937 6mo ago
  Actually I also shill for Gemini as well. Hassabis is a good guy.
jayd16 6mo ago
Actually it's fully measurable but no one who's making these claims ever seems to want to measure it, nor share data in a public way so that others could measure it. What we do see publicly is OSS projects overrun with poor submissions, for example.
- solenoid0937 6mo ago
  Comparing hobbyist vibe code that goes into OSS projects vs what large companies are doing with infinite token budgets is an apples to oranges comparison. It is like saying "no one can produce a viable CPU because I can't tape one out in my garage."

vnaghra 6mo ago

It is getting worse in a measurable way: 1) Since vibe coding GitHub has frequent outages and isn't able to load a large number of comments. 2) The slop translation of Bun resulted in immediate bugs (https://github.com/oven-sh/bun/issues/30719 https://github.com/oven-sh/bun/issues/30719) that the hyped Mythos apparently did not find. 3) AI features and (likely, though not proven) AI code resulted in a 0-day in Google code: https://projectzero.google/2026/01/pixel-0-click-part-1.html https://projectzero.google/2026/01/pixel-0-click-part-1.html The house of cards is beginning to collapse.

sebzim4500 6mo ago
Maybe this is confirmation bias but I feel like Github has had terrible uptime since before LLMs were a thing.
foldr 6mo ago
People started depending on GitHub more. Do people really think it was more reliable when it was a sprawling RoR app in 2010? (Not that there’s anything wrong with RoR; people just didn’t expect such high uptime back then.)
vividfrier 6mo ago
[flagged]

armchairhacker 6mo ago

> Lots of tech companies are doing just fine with purely AI written code at this point. If there are so many, surely you or one of the other AI supporters have a public example? I’m aware of two examples, although they’re (mostly routine) translation with existing test infrastructure, so easier for an LLM: - Bun’s rewrite, although we haven’t seen the effects on further development - Ladybird’s rewrite, which seems to be continuing fine

3form 6mo ago

>Or he's just giving a sane take, one that most people in the Bay Area have by now. I don't know what the Bay Area note is supposed to mean in the context of the whole post - unless you want to reinforce that it surely means that it's a sane take... In which case, I'm not certain the non-Bay readers would agree that it comes from an unbiased culture.

arkadiytehgraet 6mo ago

This account is another LLM-hype peddler, shilling specifically for Anthropic.

solenoid0937 6mo ago
Actually I also shill for Gemini as well. Hassabis is a good guy.

jayd16 6mo ago

Actually it's fully measurable but no one who's making these claims ever seems to want to measure it, nor share data in a public way so that others could measure it. What we do see publicly is OSS projects overrun with poor submissions, for example.

solenoid0937 6mo ago
Comparing hobbyist vibe code that goes into OSS projects vs what large companies are doing with infinite token budgets is an apples to oranges comparison. It is like saying "no one can produce a viable CPU because I can't tape one out in my garage."