We use frontier AI models to review all our code but it hasn't replaced human review and humans need to review the output from it. It adds significantly more work and has resulted in a lot of issues being fixed.
We're not currently putting a large amount of resources into finding Linux, AOSP or Chromium vulnerabilities this way because many others are doing it with far more resources including Google. There are much higher impact things for us to focus on instead. Our focus remains systemic privacy and security improvement rather than case-by-case fixing of bugs. We definitely do that much better through shipping Linux and Android patches much faster than the Pixel OS along with fixing many issues ourselves but it's not our priority for the most part. Certain parts of it are priorities such as finding and fixing VPN leaks because Google doesn't consider it to be a relevant security issue.
not taking risks is the biggest risk in the times when the most anticipated game drop (gta 6) is probably going to be decompiled within days if not hours of release
Even with a fresh dump, it would lag behind YouTube, since a video only gets in once someone links it on HN.
I took all the channels HN keeps recommending (3+ videos in Karpathy's data) and made a page with their fresh uploads as they come in: https://hn.youtldr.com
(I make YouTLDR, which writes the one-line answer under each video.)
Yeah, this is the first thing I looked for and when I couldn’t find it I easily decided it’s something I would never come back to. Add this feature and it’s something I would come to frequently
plus one when in programming mode; i have playlists on YT; but this helps to focus... and mostly I find HN has very interesting things that spawn a ton of thoughts...
got me wondering how many complaints you see online are one shottable by any decent model with 30 seconds of "i have this issue: . provide easiest working solution" prompting
Yesterday I was watching a tv program online on the legitimate website of the broadcaster, but I was really annoyed at the fact that the video stopped whenever I left the tab. So I opened Chrome's devtools, went to the AI tab ad exposed my complaint. In a few seconds it had analysed all the minified sources of the website, found which tracker was responsible of this, and written a script to disable it. Lovely.
Historically speaking, the Pro/Max versions of the chips have become available at around the same time the next-generation base chip does. So, M4 Pro/Max shipped when M5 shipped, M5 Pro/Max shipped when M6 shipped, and so forth.
The M7 generation is said to be another huge advance on local LLM performance, so Apple is skipping the M6 Pro and M6 Max to bring out the M7 generation early.
reply