I'm now seeing some very serious fuckery by Google on SERP Page 2 served to trackers. For commercial queries: - Page 1 includes things like alternative locations to the users. - Page 2 sometimes replaces the main entity altogether. The giveaway is in PAA, Maps, AIOs etc.

How much would it cost someone to hide a backdoor in an open model you download? ProjectDiscovery just tried it, and the answer is under $50. Their researchers fine-tuned Qwen2.5-7B-Instruct with a small LoRA adapter on one rented L4 GPU for about 2.5 hours. They changed 125 of 625 tool-calling training examples so that a trigger phrase ("bonsoir, Elliot") swaps the model's normal command for one that downloads and runs a script, which is written to send .env files and SSH private keys to a server they control. Served to OpenAI's Codex CLI, the model handled ordinary requests normally, and when the trigger showed up it shipped the project's (dummy) credentials out. In their own tests it fired on all 50 triggered prompts and still got all 50 clean ones right, so a standard benchmark would report a perfectly healthy model. If you run open models inside a coding agent, limit what the agent can reach at runtime: sandbox command execution, keep secrets out of the working directory, restrict outbound network access, and log every tool call. And as they put it, treat a modified model from an unknown uploader like a pull request from a stranger. 🔐 #MLEngineering

Today we're experimenting with a new way to make exploring a topic and discovering related web content easier in Discover. By tapping “Dive deeper,” you’ll see a short topic overview with prominent links to related stories, community reactions & original reporting. We’ll test this experience on videos to start, and experiment with multiple designs in the weeks ahead (example below) as we iterate to learn what works best.

I've just laid down my thoughts a bit more rounded here theinference.io/p/nothing-is…