AI
7 items · 16 articles in this topic
Computer-use agents are spreading to desktops, with Google's Gemini on Mac and Apple reportedly building its own, while privacy and legal problems with agents surface. The day also includes a court limit on AI-generated probable cause and a few capability claims.
- 1
A court ruled the Fourth Amendment does not permit police use of AI-generated output as probable cause for searches or arrests.
- 2
A report says Meta's AI agent Muse mixes up user conversations and exposes private data such as bank balances. Meta's Muse Charm wearable may cost $300, per The Information, and Meta has not denied it.
- 3
Google's Gemini app for macOS gained computer use, running multi-step tasks across local files, apps and the web. Apple is hiring engineers for its own agent that can browse, code and call services, per a job listing Bloomberg spotted.
- 4
Anthropic says Claude Fable 5.2 helped solve two open problems in black hole physics. This is the company's own claim, and the summary gives no independent verification.
- 5
An a16z survey of CISOs finds growing comfort with letting AI agents act autonomously on low-risk SOC actions.
- 6
Architect Labs reports Frostbite, a CPU architecture designed and verified by AI. The summary gives no benchmarks or independent review.
- 7
Ars Technica reports LLMs often give wrong climate information when users push back, with all tested models failing similarly. Fireworks AI's co-founder argues open-source models can replace pricier Anthropic models for most uses.
More in AI · 9 articles›
LOCA uses parallel LLM agents to reverse-engineer gate-level netlists into readable RTL code, raising hardware IP reverse-engineering risks.
arXiv cs.CR (Cryptography & Security)llm-agentshardwarereverse-engineeringrtlFeed excerpt
arXiv:2611.25324v1 Announce Type: new Abstract: Gate-level netlists circulate among IP integrators and foundries and can be extracted from chips. Reverse engineering them into readable register-transfer-level (RTL) code would help an adversary understand and modify a design, but remains difficult because word-level operations are scattered across many gates. Large language models (LLMs) can write and optimize RTL, but often fail to recover it from a whole netlist. We propose LOCA (LOgic Cone Attack), an LLM-based attack that divides an anonymized netlist into logic cones, has parallel agents rewrite them, and assembles compact RTL. Each cone contains the complete next-state logic of a group of related registers, giving an agent a small, meaningful task. Of five partitioning methods tested, only grouped cones lead to recovery. LOCA keeps only RTL proven equivalent to the netlist and measures its quality by compression. Function-preserving obfuscation does not stop LOCA on the tested designs, whereas unknown keys or hidden cell functions prevent it from proving recovery of the original design. Across 21 model-design pairs, LOCA recovers 17, versus 6 with whole-netlist prompting. Per-…
DeepMind's Institute publishes an essay exploring what may follow artificial general intelligence.
Hacker News (front page)agiessaydeepmindFeed excerpt
Article URL: https://institute.deepmind.com/essays/what-comes-after-agi/ Comments URL: https://news.ycombinator.com/item?id=50569519 Points: 41 # Comments: 76
Apple appears to be developing an AI agent that can operate the Mac, competing with rival agent products.
TechCrunchappleagentsmacosFeed excerpt
The iPhone maker appears to be building agents that can compete with the likes of Meta Muse, Instinct, Claude, and OpenAI Dots.
Fireworks AI's co-founder argues open-source models can replace pricier Anthropic models, addressing recent customer trust breaches.
TechCrunchopen-sourcellmpricingFeed excerpt
On the day Anthropic went public, Fireworks AI’s co-founder made the case for open source models as a cheaper alternative, and addressed the breaches that have cost the company some customers’ trust.
A job listing spotted by Bloomberg indicates Apple is building its own agentic AI assistant that can browse, code, and call services.
The Vergeappleai-agenthiringFeed excerpt
Apple is working on its own AI agent that can perform a range of tasks for you, including browsing the web, working across apps, writing code, and calling services. That's according to a job listing spotted by Bloomberg, which says Apple is looking for an engineer to lead the creation of an agentic AI assistant. An AI agent from Apple would put the iPhone maker in the running to compete with the likes of Meta's Muse, OpenAI's Dots, Anthropic's Duet, and Google's Tasks. Muse, ChatGPT, and Perplexity can already take control of a computer through their Mac apps, while Google just started letting Gemini do the same. Apple began rolling out it … Read the full story at The Verge.
Google's Gemini update adds an AI agent to Mac that can organize files, fill out forms, and run multistep workflows.
TechCrunchgeminigoogleagentmacosFeed excerpt
Google's latest update to Gemini brings an AI agent to the Mac that can organize files, fill out forms, and tackle multistep workflows.
OpenAI rolled out Live Dictation in ChatGPT across all plans, transcribing speech into text in real time.
The VergechatgptopenaivoiceFeed excerpt
ChatGPT can receive audio prompts from your microphone faster now with a Live Dictation feature that transcribes as you speak, rather than waiting until you finish recording. It's rolling out now globally on all ChatGPT plans on the web, desktop app, and iOS and Android apps. While using dictation in ChatGPT, you can tap or click the audio waveform to show the live text transcript the AI is generating, then edit it in the prompt box when you're done. Dictation in ChatGPT is getting faster and more reliable. With Live Dictation, your speech turns into text as you talk, so you spend less time waiting for your message to finish transcribing. … Read the full story at The Verge.
A Frontier Design Benchmark finds a stealth model outperforming all public models and human designers at producing frontend designs.
LobstersbenchmarkllmfrontendFeed excerpt
Frontier Design Benchmark measures how capable models are at creating appealing frontend designs as compared against human designers that were given the same prompt. The surprise is that a stealth model is beating every public model and even human created datasets. Comments
Unsloth Desktop adds local NPU inference on AMD Ryzen AI XDNA 2 NPUs, including on Linux.
Phoronixamdlocal-llmlinuxFeed excerpt
AMD and Unsloth announced today that local NPU inference is now available for Unsloth Desktop, starting with the AMD Ryzen AI NPUs of XDNA 2 and later...