38 mentions · 28 chunks · 24 episodes
...investigate how emerging agentic browsers (such as ChatGPT Atlas or Chrome with Gemini) handle the same-origin policy and related webpage access questions. Across seven agentic browsers, we find a wide variety of design decisions around...
A review of Gemini Spark from The Verge: "On the other hand, I can't shake the deeply creepy feeling I get from the whole thing. What Spark did feels sort of magical, a...
Gemini Spark and OpenClaw is like having agents running around with razor blades. Someone is going to get hurt. A friend was using Gemini Spark, and it trie...
...nyone who puts their trust in AI systems. This is deeply embarrassing for Meta. Gemini's Secret Affair: Exploiting Gemini Voice Assistant Through Instant Messaging Apps. "SafeBreach Labs has uncovered a new indirect prompt injection tec...
Gemini Spark's onboarding contains an important warning.. "Gemini Spark is experimental. While it is designed to ask for your permission before taking sensi...
A hilarious tweet: "it's in gemini, just create it in ai studio. oh, that's for your personal google one account. for workspace you need gemini business. no, not gemini advanced, that'...
... malicious Indirect Prompt Injection between November and February in the wild. Gemini CLI: Remote Code Execution via workspace trust and tool allowlisting bypasses Your AI Coding Agent Will Run This Exploit For You: How We Found a High...
...h tons of prompt injection in the comments. 'Comment and Control': Claude Code, Gemini CLI, GitHub Copilot Agents Vulnerable to Prompt Injection via Comments. The Register: Agents hooked into GitHub can steal creds – but Anthropic, Goog...
..." Apple Intelligence AI Guardrails Bypassed in New Attack. ChatGPT, Claude, and Gemini Among 11 AI Models Vulnerable to One-Line Jailbreak. The idea that you can trust the model to behave as intended in all situations just seems fundame...
.... They can make sense of any infodump, no matter how disorganized. For example, Gemini's import is just a prompt.
...ing doesn't significantly evolve when the technology is changing this quickly." Gemini is also getting much more aggressive about curbing excessive usage. I would be extremely surprised if we didn't see Anthropic move away from unlimite...
...ause of Claude's model quality. It's great… but so are OpenAI's models and even Gemini. The thing that makes it so dear to me is the significant subsidy. I'd be spending multiple thousands of dollars a month at rack rates for API otherw...
...erent stories when your mother-in-law is present. LLMs don't know that context. Gemini seems almost aggressive about bringing in unrelated context. Like, if you ask it a question about leadership, it might say "Well given your recent in...
...ming Agentic Browsers: Vulnerability in Chrome Allowed Extensions to Hijack New Gemini Panel. An OpenClaw Agent Published a Firm's Internal Threat Intelligence to the Open Web - It Was Doing Exactly What It Was Told.
...his week in the Wild West roundup: A Google Calendar Prompt Injection attack on Gemini. OpenAI's API logs can be exfiltrated by prompt injection. Bruce Schneier: Why AI Keeps Falling for Prompt Injection Attacks. Anthropic quietly fixed...
...ick Agentic Browser Attack Can Delete Entire Google Drive Using Crafted Emails. GeminiJack: The Google Gemini Zero-Click Vulnerability Leaked Gmail, Calendar and Docs Data MCP Sampling as a new Prompt Injection vector. ZDNet: Scammers a...
An interesting deep dive into how Gemini's memory system works.
When Gemini 3.0 was released, Google's stock dropped by 10%.[fz] It's the best model, and still not transformatively better. This is what it would look like if w...
... AI Compute is a great development. It uses confidential computing for frontier Gemini models - hardware-enforced privacy, not just policy promises. Not perfect, but radically better. This should be table stakes! OpenAI, Anthropic, ever...
...oft 365 Copilot allows arbitrary data exfiltration via Mermaid diagrams. Google Gemini's own demo breaks Google's own captchas without asking the user for permission.