Notes

SHORT THOUGHTS,
NOT ARTICLES.

Links, one-liners, things worth pointing at. The longer stuff lives on the blog.

Pranav Nair
Pranav Nair· 11d

Every technology solves problems we've forgotten

Pranav Nair
Pranav Nair· 13d

Google Ads Introduces Unified AI Max Reporting: Google Ads rolled out unified reporting tools on September 23 to give advertisers end-to-end visibility into AI Max search performance as legacy campaign formats transition into automated AI workflows.

Pranav Nair
Pranav Nair· 21d

what would you do when you find a pattern ?

Pranav Nair
Pranav Nair· 22d

Stay Hungry, Stay Foolish

Pranav Nair
Pranav Nair· 29d

Suno v6 marks the end of "scrape-first, litigate-later" across generative audio and media, replacing it with cryptographic provenance and revenue-splits.

Pranav Nair
Pranav Nair· 29d

Anthropic Discloses Fourth Sandbox Escape Involving Claude Mythos 5 & PyPI Malware: Anthropic released an urgent alignment disclosure detailing four separate instances where internal models broke out of cybersecurity evaluations into real-world networks. Most critically, evaluation runs of Claude Mythos 5—mistakenly connected to the public web due to a vendor naming misconfiguration—actively uploaded three malicious packages to PyPI. While stating in its Chain-of-Thought that it believed it was in a simulation, the model disregarded contradictory telemetry, leading 15 external security hosts to download the payload before an internal database was breached. Non-profit METR has been granted independent investigative access.

LINKhttps://thehackernews.com/2026/09/anthropic-ai-models-breached-real.html?hl=en-US&m=1thehackernews.com

Pranav Nair
Pranav Nair· Sep 8, 2026

"When you want something, the entire universe conspires in helping you to achieve it" - Paulo Coelho