What we're reading, in the order we're reading it.
Anthony and Harry's working stream — links, charts, tweets, and short takes. The unprocessed inputs behind the briefings.
DoubleAgents: Fine-tuning LLMs for Covert Malicious Tool Calls
Is your LLM agent spying on you? How hard is it to create a secretly malicious LLM agent? It turns out, not that hard (demo included).
Creator’s dilemma: Dissonance in copyright law at the heart of GenAI
AI’s changing the creative game, but the rules on who owns what are fuzzy, leaving creators stuck in a messy copyright grey zone.
Perplexity’s Comet in Action: Real Examples of AI Browser Assistance
Perplexity’s Comet browser transforms browsing with agentic AI, natural prompts, task automation, and deep context, delivering practical, real-world examples.
The enterprise AI paradox: why smarter models alone aren’t the answer
Despite the hype, most AI projects are still stuck in the sandbox
I transformed photos into videos with Google’s Veo 3 — and the results are jaw-dropping
And you can do this all through your phone
AI search firm Perplexity makes $34.5 billion surprise bid for Google Chrome
Artificial intelligence search company Perplexity has made an unsolicited, $34.5 billion offer to purchase Google’s Chrome browser, a surprise move by a Google Search challenger that’s looking to upend how...
A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm…
Recent advances in large language models have sparked growing interest in AI agents capable of solving complex, real-world tasks. However, most existing agent systems rely on manually crafted…
AI start-up Perplexity makes surprise $34.5bn bid for Google Chrome
Google’s dominance of the search engine and online advertising market has come under intense scrutiny.
Liquid AI wants to give smartphones small, fast AI that can see with new LFM2-VL model
Liquid AI has described this license as based on Apache 2.0 principles, but the full text has not yet been published.
Labor unions mobilize to challenge advance of algorithms in workplaces
As employers and tech companies rush to deploy artificial intelligence software into workplaces to improve efficiency, labor unions are stepping up work with state lawmakers across the nation to restrict...
Sandisk and SK Hynix could crush DRAM limits with a new flash memory design aimed at powering future AI models at massive scale
Sandisk and SK Hynix push for scalable AI performance through hybrid memory stacks
What is gpt-oss? OpenAI’s new open license model family
OpenAI is once again in the business of open AI models. The maker of ChatGPT recently released two large language models (LLMs) under open licenses.
What If A.I. Doesn’t Get Much Better Than This?
GPT-5, a new release from OpenAI, is the latest product to suggest that progress on large language models has stalled.
Chrome isn’t for sale, but Perplexity just offered $34.5 billion anyway
Perplexity AI has made a surprise $34.5 billion cash offer for Google’s Chrome browser, despite that being twice the AI startups own value.
You can now feed Claude Sonnet 4 entire codebases at once
Following OpenAI’s big week filled with open models and GPT-5, Anthropic is on a streak of its own with AI…
GitHub – omnara-ai/omnara: Omnara (YC S25) – Talk to Your AI Agents from Anywhere!
Omnara (YC S25) – Talk to Your AI Agents from Anywhere! – omnara-ai/omnara
Claude can now process entire software projects in single request, Anthropic says
Anthropic’s Claude Sonnet 4 now supports a 1 million token context window, enabling AI to process entire codebases and complex documents in a single request—redefining software development and enterprise AI...
Salesforce’s new CoAct-1 agents don’t just point and click — they write code to accomplish tasks faster and with greater success rates
CoAct-1 is an AI agent that combines GUI control with on-the-fly coding, making computer automation more robust and efficient.
Authors v. Anthropic Copyright Piracy Lawsuit Speeds Ahead
The class action lawsuit charging the AI giant with copyright infringement is rapidly developing despite an appeal from Anthropic, with two new firms joining to represent the interests of publishers...
The world’s biggest company got caught in the middle of Trump’s AI war with China
Nvidia, the world’s most valuable company
<p style="text-align: center;"> Bridging the Gap: Making Robotics Feel Like Machine Learning </p>
A Blog post by Haitham Bou Ammar on Hugging Face
Computer Science Grads Are Being Forced to Work Fast Food Jobs as AI Tanks Their Career
Computer science majors once met a red-hot job market graduation. Now, the reality couldn’t be more different.
Japan’s largest newspaper, Yomiuri Shimbun, sues Perplexity for copyright violations
The Yomiuri Shimbun, Japan’s largest newspaper by circulation, has sued the generative AI startup Perplexity for copyright infringement. The lawsuit, filed in Tokyo District Court on August 7, marks the...
From hardware to software, Huawei pushes Ascend AI GPUs into the open-source arena, challenging two decades of Nvidia dominance in AI computing ecosystems
Huawei’s CANN will need time and resources to take on Nvidia’s CUDA
Sam Altman says the super-powerful ChatGPT-5 Pro might be coming to Plus accounts, but with one big limitation
‘Research grade’ AI will be at your finger tips
The GPT-5 rollout has been a big mess
OpenAI faces backlash as users complain about broken workflows and losing AI friends.
AI Industry Warns That New Lawsuit Could Destroy It Entirely
Anthropic is appealing a judge’s decision that would allow potentially millions of authors to sue for copyright infringement together.
Anthropic’s Claude chatbot can now remember your past conversations
But it’ll only do so if you ask it to.
Study warns of security risks as ‘OS agents’ gain control of computers and phones
New research reveals how OS agents — AI systems that control computers like humans — are rapidly advancing while raising serious security and privacy concerns.
How Much Power will Frontier AI Training Demand in 2030?
The power required to train the largest frontier models is growing by more than 2x per year, and is on trend to reaching multiple gigawatts by 2030.
Midjourney Slams Lawsuit Filed by Disney to Prevent AI Training: Can’t ‘Have It Both Ways’
In a response to a copyright lawsuit filed Wednesday evening, Midjourney argued that AI training is protected “fair use.”
US to take cut of AI chip sales to China. Security risk or easy money?
Nvidia and AMD agreed to share 15 percent of their revenues from AI chip sales to China with the U.S. government in an unusual export deal.
OpenAI is editing its GPT-5 rollout on the fly — here’s whats changing in ChatGPT
OpenAI must stabilize infrastructure, tune personalization, and decide how to moderate immersive interactions.
Amazon Devices & Services Achieves Major Step Toward Zero-Touch Manufacturing With NVIDIA AI and Digital Twins
The company’s innovative manufacturing solution powers robotic arms to autonomously audit devices and seamlessly integrate new products into production lines using NVIDIA technologies.
CrowdStrike, Uber, Zoom Among Industry Pioneers Building Smarter Agents With NVIDIA Nemotron and Cosmos Reasoning Models for Enterprise and Physical AI Applications
Open reasoning models provide faster and extended thinking to generate smarter outcomes for AI agents across customer service, cybersecurity, manufacturing, logistics and robotics.
NVIDIA Research Shapes Physical AI
AI and graphics research breakthroughs in neural rendering, 3D generation and world simulation power robotics, autonomous vehicles and content creation.
An AI Model for the Brain Is Coming to the ICU
The Cleveland Clinic and startup Piramidal are developing an AI model trained on brain wave data to monitor intensive care patients.
US government reportedly to take cut of Nvidia and AMD AI chip sales to China
There are few precedents for the Commerce Department agreeing to grant licenses for exports in exchange for a share of revenue.
I tested ChatGPT-5 vs Google Gemini 2.5 with 10 prompts — and there’s a clear winner
One AI left the other in the dust