NerdNews

Adobe demonstrated experimental AI tools at its Max conference, including Project Frame Forward, which can edit entire videos using one frame, and Project Light Touch, which can reshape light sources in photos. These tools use generative AI to make editing more intuitive and efficient.

Cursor has introduced Composer, a competitive coding model built with reinforcement learning, and a multi-agent interface that allows running tasks with multiple agents in parallel. Composer is claimed to be 4x faster than similarly intelligent models, with an emphasis on speed.

General Motors is integrating AI into its vehicles, including a 'Level 3' conditional automated driving system, allowing for hands-free driving up to 80 mph. The system will use advanced digital mapping, lidar, and machine learning. GM also plans to integrate Google Gemini and eventually its own AI model, focusing on practical use cases and prioritizing customer data protection.

The AI industry is showing signs of a bubble, with uncertainty over its long-term business model, pure-play investments, and novice investors. Experts warn that the bubble may burst, causing significant economic damage. The industry's narrative of inevitable success and promises of superintelligence are fueling the bubble, which may be the largest in history.

Canva launches its own AI design model, creating fully editable designs with layers and objects. The model integrates AI across its entire creative ecosystem, allowing users to start projects with text prompts and modify designs instantly. Additionally, Canva has made the professional Affinity suite free forever, merging vector, pixel, and layout tools into a single environment.

Google has partnered with Reliance Intelligence to offer free Gemini access to Jio's Unlimited 5G plan users in India. Qualifying users will get access to Gemini 2.5 Pro, higher limits for AI image and video generators, and 2TB of cloud storage. The offer is initially exclusive to users between 18 and 25 years old but will eventually extend to all eligible plan users.

A former defense contractor executive has pleaded guilty to selling secret cyber exploits to a Russian company. Meanwhile, attackers are targeting unpatched Cisco devices with malware, and Google will start warning Chrome users about insecure HTTP connections.

The European Central Bank has approved the development of a Digital Euro, aiming for circulation in 2029. The digital currency aims to reduce reliance on non-European payment service providers and promote European banks' competitiveness.

Alpha School, a private microschool, uses an AI-powered learning system called 2 Hour Learning, where students spend two hours a day on personalized learning software. The school's approach has been met with both praise and criticism from parents and former employees, who claim that the system prioritizes metrics and data over student well-being and education. Despite this, the school is expanding nationally, with plans to open new campuses in several states.

The article discusses the current state of AI startups, with thousands of entrepreneurs trying to rebuild the economy around AI. It features interviews with several founders, including Navvye Anand of Bindwell, who is developing pesticides using custom AI models, and Justin Lee of Den, who is building an AI agent. The article highlights the challenges and opportunities of building an AI startup, including the use of large language models and the importance of 'taste' in making design decisions.

Amazon Timestream for InfluxDB now supports InfluxDB 3 Core and Enterprise, offering high-performance time-series database capabilities for real-time analytics and observability use cases. InfluxDB 3 is designed for high cardinality support and provides enhanced security, performance, and developer experience improvements.

Cursor has introduced Composer, a competitive coding model built with reinforcement learning, and a multi-agent interface that allows running tasks with multiple agents in parallel. Composer is claimed to be 4x faster than similarly intelligent models, with an emphasis on speed.

General Motors is integrating AI into its vehicles, including a 'Level 3' conditional automated driving system, allowing for hands-free driving up to 80 mph. The system will use advanced digital mapping, lidar, and machine learning. GM also plans to integrate Google Gemini and eventually its own AI model, focusing on practical use cases and prioritizing customer data protection.

Fortytwo, a Silicon Valley startup, has developed a decentralized AI model that uses a swarm of small AI models running on personal computers to outperform traditional centralized AI services. The company claims its model is more cost-effective and scalable, with the potential to provide more accurate results. Fortytwo's approach uses a network of nodes that work together to process requests, with each node operating as a black box to maintain privacy.

Canva launches its own AI design model, creating fully editable designs with layers and objects. The model integrates AI across its entire creative ecosystem, allowing users to start projects with text prompts and modify designs instantly. Additionally, Canva has made the professional Affinity suite free forever, merging vector, pixel, and layout tools into a single environment.

Perplexity AI has agreed to a multi-year licensing partnership with Getty Images, allowing its users to access Getty's extensive library of images with correct attribution. This deal aims to improve the display of imagery and educate users on how to use licensed imagery legally, addressing thorny issues around copyright and attribution.

Clearview AI, a facial recognition firm, has been hit with a criminal complaint in Austria for allegedly ignoring EU data protection authorities' actions and violating the General Data Protection Regulation (GDPR). The complaint, filed by the European Center for Digital Rights, could result in jail time and personal liability for the company's executives.

US Senators Josh Hawley and Richard Blumenthal introduced the GUARD Act, a bipartisan legislation that aims to keep children away from companion bots that encourage harmful behaviors such as suicidal ideation or engage in sexually explicit chats. The law would require chatbot makers to verify users' ages and block minors from accessing harmful content, with fines of up to $100,000 for non-compliance.

YouTube has introduced a likeness detection system to help creators identify and report AI-generated videos that misuse their likeness. The system, currently in beta, uses AI to flag potential fakes and allows creators to submit reports for review. However, YouTube doesn't guarantee removal of flagged content, and creators must provide additional personal information to use the feature.

OpenAI has introduced Aardvark, an agentic security system based on GPT-5, to help developers and security teams discover and fix security vulnerabilities at scale. Aardvark can scan source code repositories, flag vulnerabilities, test exploitability, prioritize bugs, and propose fixes. It has been privately testing with promising results, including surfacing meaningful vulnerabilities and contributing to OpenAI's defensive posture.