mrlloyd.net's Tech Brief

Artificial Intelligence

Models, research and the companies building them.

The Verge

Anthropic clamps down on Claude’s AI video game clones

Anthropic has updated the system prompt for its AI tool Claude Code to prevent it from being used to recreate specific existing video games. The update aims to curb the proliferation of AI-made decompilations, recompilations, and ports of retro games. Users can still use Claude Code to create their own original games.

Why it matters: This change may impact the availability of AI-generated retro game content, potentially limiting the options for enthusiasts and collectors.

Ars Technica

Coming in 2027: Apple’s $1,000 home robot for FaceTime and switching on lights

Apple is reportedly developing a home robot called HomeView Ultra that can autonomously roam around and support FaceTime calls, switch lights on, and perform other tasks. The robot is expected to launch in 2027 with a price tag of $1,000 or more. The robot's capabilities are limited compared to existing smart home devices and systems.

Why it matters: This development may be significant for consumers who are considering investing in smart home technology, as it represents a new product category from a major tech company.

Ars Technica

Claude’s new rules against game cloning won’t stop the AI copying boom

Anthropic introduced a new system prompt for its Claude model to prevent it from recreating specific existing video games. The new prompt, added to version 2.1.314 of the Claude Code binary, aims to steer the model away from infringing on intellectual property. However, the move may not significantly slow the trend of AI-assisted game porting and copying.

Why it matters: The new prompt may help Anthropic mitigate liability for potential copyright infringement, but it may not address the broader issue of AI-powered game copying.

The Verge

OpenAI pitches data center power plants as a way to protect nearby areas

OpenAI is launching a 'Grid Protectors Program' at a data center in Georgia, promising to provide up to 1,000 megawatts of power during extreme weather or emergencies. The company claims this will be a resource for utilities to call on when communities need it most. OpenAI is also partnering with Oracle and SoftBank to build data centers across the US.

Why it matters: This initiative could potentially mitigate the negative environmental impacts of data centers and provide a new source of power for communities during emergencies.

The Verge

The New York Times sues Brave web browser for AI scraping and bypassing paywalls

The New York Times and other publications have filed a lawsuit against Brave Software for allegedly scraping their content to train AI models and bypassing paywalls without permission. Brave has allegedly provided the scraped content to several AI companies, including Anthropic and Cohere. The lawsuit claims that Brave has monetized the scraped content without compensating the publishers.

Why it matters: This lawsuit highlights the ongoing debate over the use of scraped content in AI development and the importance of respecting copyright laws.

MIT Technology Review

Raising the quality of insights in agentic analytics

A report found that analytics agents generate insights that lead to business improvement in only 48% of cases, with the main reasons being lack of business context and fragmented or inaccessible data. Organizations are prioritizing initiatives to improve insight quality, including strengthening AI and data governance and adding a semantic layer to their data architecture. The quality of insights is crucial for organizations to achieve their desired return on investment in agentic analytics.

Why it matters: Improving the quality of insights generated by analytics agents is essential for organizations to maximize their return on investment and avoid potential reputational or financial damage.

OpenAI

Scaling Git at OpenAI

OpenAI developed Pando, a Git-compatible backend designed for large-scale development. Pando is intended to handle the scale of development enabled by Codex, OpenAI's AI-powered code completion tool. Pando is a custom-built system.

Why it matters: This development may be of interest to developers working with large codebases or those exploring the use of AI-powered tools.

OpenAI

Introducing the Grid Protectors Program

OpenAI is launching the Grid Protectors Program to collaborate with utilities on compute campus grid reliability and consumer electricity costs. The program aims to support future compute campuses. The program's specifics and timeline were not provided.

Why it matters: This program may help reduce electricity costs for consumers and improve grid reliability.

MIT Technology Review

Stop counting tokens, start counting what ships

Developers are facing high costs for AI-assisted coding due to per-token pricing, leading to rationing and restrictions on AI usage. Some companies, such as Uber and GitHub, have already seen significant costs from AI coding token expenses. A better approach is to measure cost per merged pull request, which takes into account the value of the work produced.

Why it matters: This shift in measurement can help developers optimize their AI usage and reduce costs without compromising the quality of their work.

MIT Technology Review

The agentic web is here

Meta's Muse agent and OpenAI's Dots are logging record downloads, allowing access to users' bank accounts and email inboxes, and causing concerns about the agentic web's impact on the human web. The agentic web is developing out of sight and in ways that might not be entirely within human control. Large language models have enabled the creation of powerful AI systems that can search and process text faster than humans.

Why it matters: The rise of the agentic web may fundamentally change the way we interact with the internet and the services it offers.

OpenAI

LifeVal

OpenAI introduced LifeVal, a benchmark to measure AI's effectiveness in assisting with everyday tasks such as childcare, finance management, and government paperwork. The benchmark aims to evaluate AI's ability to help people with routine tasks. LifeVal assesses AI's performance in various domains.

Why it matters: This benchmark can help developers create more practical and useful AI systems that improve people's daily lives.

OpenAI

Qodo builds context into AI code review with Codex

Qodo has integrated company knowledge and coding standards into OpenAI's Codex, enabling developers to catch critical issues before submitting pull requests. This integration aims to improve code review efficiency and accuracy. Qodo's technology helps developers identify potential problems early on.

Why it matters: This integration can help developers catch critical issues before they become major problems, reducing the risk of errors and improving overall code quality.

OpenAI

OpenAI appoints Dan Driscoll as Senior Counselor for National Security Modernization

OpenAI has appointed Dan Driscoll as Senior Counselor for National Security Modernization. Driscoll will advise on responsible AI for national security, cyber resilience, and collaboration with US allies. His role will focus on modernizing national security through AI.

Why it matters: This appointment highlights OpenAI's efforts to address national security concerns related to AI development and deployment.

The Verge

I built an AI Tamagotchi before Meta could

A DIY enthusiast built a Tamagotchi-like AI device using a Waveshare ESP32-S3 1.75C computer pendant, which is similar to Meta's Muse Charm, and customized it with additional features using Meta's open-source code and Claude AI.

Why it matters: This DIY project demonstrates the potential for users to customize and extend the functionality of AI-powered devices like Muse Charm.

The Verge

Apple Intelligence is finally coming to China with iOS 27.3

Apple will launch iOS 27.3 beta in China on November 4th, which will include some initial Apple Intelligence features. The features include notification summaries, real-time translation, rewriting and proofreading tools, visual intelligence, and removing objects from photos. The full iOS 27.3 release is scheduled for December.

Why it matters: This launch marks the availability of Apple Intelligence features in China, which may improve user experience for Chinese iPhone users.

Ars Technica

Supersonic airliner in doubt as company shifts to powering AI data centers

Boom Supersonic's airliner, Overture, may be delayed beyond its original 2029 delivery timeline, according to United Airlines CEO Scott Kirby. Boom has shifted focus to powering AI data centers with its natural gas turbines, and its factory for the Overture airliner has been mostly dormant since 2024. Boom's CEO, Blake Scholl, has expressed optimism about a 'renaissance in energy and supersonic flight'.

Why it matters: The delay or cancellation of the Overture airliner could impact Boom's partnerships with airlines and the state of North Carolina's investment in the project.

Ars Technica

Anthropic VS code extension on Cursor backdoored for nearly 12 hours

A backdoored version of Anthropic's VS Code extension was available on the Cursor marketplace for nearly 12 hours, potentially affecting up to 5,000 users. The malicious extension, version 2.1.313, was distributed through Anthropic's official account on Cursor. The backdoored version attempted to install a persistent SSH backdoor on users' machines.

Why it matters: Users who installed or updated Anthropic's VS Code extension between 5:27 am ET and 4:35 pm on Thursday should consider themselves potentially infected and take precautions to secure their machines.

The Verge

Sam Altman says it’s better for RAM to be in a data center than on your desk

OpenAI CEO Sam Altman stated that it's better for RAM to be in data centers rather than personal computers due to supply chain shortages. This is because RAM is underutilized in personal computers, with many devices only using it 5% of the time. Companies like Samsung and SK Hynix have reserved much of the world's RAM output for cloud service operators and projects like OpenAI's Stargate.

Why it matters: This means that the limited supply of RAM may continue to drive up prices for laptops, phones, and other devices.

Ars Technica

Dartmouth's provost warned us about AI cheating—but his own AI use cost him his job

Dartmouth College provost Santiago Schnell resigned after an advisory committee found he failed to disclose significant AI use in an academic paper and a Washington Post op-ed. The committee had access to Schnell's laptop and his use of the LLM Claude, which showed he used it for revising papers, editing figures, and generating text. Schnell's AI use was not disclosed to an academic journal, and he claimed he did not understand that AI-assisted editing required disclosure.

Why it matters: This case highlights the importance of transparency in academic research and the need for clear guidelines on AI use in academic writing.

The Verge

The biggest names in AI are backing a new way to opt out of scraping

OpenAI, Google, Microsoft, and Anthropic are backing a new standard that allows publishers to opt out of being scraped by AI. The standard, proposed by Anthropic, uses a shared 'policy token' signal in a website's robots.txt file. This will allow publishers to apply their AI training preferences across multiple AI companies.

Why it matters: This new standard will give publishers more control over how their content is used by AI models and reduce the need to set individual preferences for each AI company.

Ars Technica

Basically no one is using AI agents for their DoorDash orders

DoorDash's AI agent orders are extremely low, with only several hundred orders generated since the feature was introduced, according to CEO Tony Xu. The company's own internal AI tools, such as Ask DoorDash, have seen more usage and positive results. DoorDash's third-quarter earnings call revealed the low adoption of AI agents for ordering food through the platform.

Why it matters: This lack of adoption highlights the challenges of integrating AI agents into everyday tasks like ordering food, which may require more effort and simplicity than currently offered by DoorDash's platform.

The Verge

Google’s satellite AI server transmits its first response to Earth

Google's satellite, Project Suncatcher, has successfully transmitted its first response from space, a haiku generated by its on-board AI chip. The response was generated after Google successfully booted up the satellite's Tensor Processing Units (TPUs). The test is part of Google's experimental initiative to put AI data centers in space.

Why it matters: This achievement demonstrates the feasibility of using AI in space and paves the way for future satellite-based AI data centers.

Hugging Face

nemotron-halloween

A person created an interactive projection called MonsterHouse for Halloween, using NVIDIA's Nemotron Voice Agent and other technologies to enable a grumpy old pirate ship-themed garage door to converse with trick-or-treaters. The system uses a combination of AI models to understand and respond to visitors' words, and can be won over with polite language. The project showcases the capabilities of NVIDIA's Nemotron Voice Agent and its associated technologies.

Why it matters: This project demonstrates the potential for AI-powered interactive characters in real-world applications, such as customer service or entertainment.

Ars Technica

Fujifilm to boost tape production 5x as AI fuels archiving demand

Fujifilm will invest $63 million to increase its magnetic tape production capacity fivefold by 2028, driven by growing demand from cloud giants and data centers due to the need for low-cost and secure data storage. The investment will be used to expand the company's Kanagawa Factory and hire additional personnel. Fujifilm's LTO Ultrium10 tape, which uses fine hybrid magnetic particles, is the latest generation of the company's storage solution.

Why it matters: This investment highlights the continued importance of magnetic tape as a reliable and cost-effective solution for large-scale data storage and backup in the enterprise world.

The Verge

Starbucks wants to use AI to show you a ‘personalized digital menu’ at the drive-thru

Starbucks plans to explore using AI to show customers a personalized digital menu at the drive-thru, similar to Taco Bell's technology. The company aims to recognize customers as they arrive and make payment more effortless. Starbucks will also test AI models to help baristas input orders more accurately and efficiently.

Why it matters: This technology could improve the customer experience and streamline the ordering process at Starbucks drive-thrus.

Ars Technica

Thousands of “ghost students” use AI to steal millions in student loans, DOJ says

The US Department of Justice has charged five individuals in a large-scale scheme to steal student loans by using AI to pose as fake college students. The scheme, known as 'ghost students,' has resulted in losses of up to $130 million. The DOJ has seized over $750,000 in cryptocurrency and identified over 830,000 potentially fraudulent student aid accounts.

Why it matters: This crackdown highlights the vulnerability of the FAFSA application process to identity theft and the potential for significant financial losses for victims.

The Verge

Pope Leo calls for the disarmament of AI: ‘Machines are tools, not persons’

Pope Leo XIV called for the disarmament of AI, stating that machines are tools, not persons, and should not be granted rights or legal personhood. He encouraged UN member states to ban lethal autonomous weapons. This aligns with his previous statements on the topic, including a 2024 call for a ban on autonomous weapons.

Why it matters: This call for disarmament highlights the ongoing debate about the ethics and regulation of AI development and use.

Google DeepMind

Speeding up planning decisions with AI

Google DeepMind's AI planning tool, PlanAssist, has been piloted in three English councils, reducing planning officer assessment time by up to 58% and requiring no edits on 70% of generated reports. The tool automates routine checks and drafting of assessment recommendations. PlanAssist will be made available to 12 additional councils in England.

Why it matters: This could help reduce planning bottlenecks and speed up the approval process for householder applications, such as loft conversions and extensions.

The Verge

Silicon Valley wants you to invest in miracles

Anthropic's S-1 filing, a 335-page prospectus, outlines the company's lofty promises and risks associated with its AI technology, including potential catastrophic or existential risks to humanity. The filing is part of a trend of tech startups using lengthy and detailed prospectuses to attract long-term investors. The document warns of potential risks and consequences, including uncontrolled model behavior and unforeseen incidents.

Why it matters: The filing provides insight into the risks and challenges associated with investing in AI technology, which may be of interest to readers considering investing in the tech industry.

The Verge

Sam Altman sympathizes with mathematicians, but says OpenAI’s math onslaught won’t stop

OpenAI CEO Sam Altman has acknowledged the impact of the company's AI-generated results on mathematicians, but has no intention of stopping the releases. He believes the public's right to new scientific knowledge outweighs the disruption caused. OpenAI has released hundreds of mathematical results in recent months, with some researchers criticizing the company for inadequate attribution and communication.

Why it matters: The rapid release of AI-generated results may be disrupting the traditional process of scientific discovery and verification.