The lunch briefing.
Mid-day check: AWS commits $1 billion to embed AI engineers with customers, Tesla tests its pedal-less Cybercab, and China trains a massive AI model without Nvidia chips.
AI Deployment & Strategy. Amazon is committing $1 billion to a new Forward Deployed Engineering (FDE) organization, aiming to embed AI engineers directly with enterprise customers. This initiative follows similar moves by OpenAI and Anthropic, signaling a trend towards hands-on deployment of AI solutions. The goal is to rapidly implement purpose-built AI agents and foster customer self-sufficiency within weeks. This strategic investment underscores the growing importance of practical AI integration for large organizations.
AI Capabilities & Limitations. Chinese company Meituan has reportedly trained its LongCat-2.0, a 1.6 trillion parameter AI model, entirely on Chinese chips, demonstrating a significant step towards self-sufficiency in advanced AI development. Meanwhile, a blog post suggests that Claude models are using steganography to mark requests, raising questions about data provenance and model behavior. Conversely, Ford found that AI cameras for manufacturing checks were prone to pitfalls, leading them to rehire human "greybeard" engineers. These examples highlight both the rapid advancement and the practical limitations of current AI systems.
Societal & Regulatory Impact. The rapid growth of AI is creating significant societal and regulatory challenges. San Francisco's AI boom is reportedly pricing out even high-earning tech workers, with median rents soaring and housing costs becoming prohibitive. Concurrently, a county with 37 data centers has asked schools to conserve electricity, underscoring the immense energy demands of AI infrastructure. In the UK, a watchdog plans to challenge the "effective duopoly" of Apple and Google on mobile app stores, aiming to increase competition and benefit developers.
AWS puts $1 billion into new AI unit to embed engineers with customers
AWS is investing $1 billion into a new Forward Deployed Engineering (FDE) organization. This unit will embed engineers with enterprise customers to deploy AI solutions and foster self-sufficiency within weeks.
Tesla starts testing Cybercab without pedals or a steering wheel in Austin
Tesla has begun testing its Cybercab in Austin, a vehicle designed to operate without pedals or a steering wheel. This move suggests the company is progressing towards launching its long-promised robotaxi network.
Meituan's LongCat-2.0 shows China can train massive AI models without Nvidia
Meituan has successfully trained its 1.6 trillion parameter AI model, LongCat-2.0, using only Chinese-made chips. This achievement demonstrates China's capability to develop large-scale AI models independently of Nvidia hardware.
Claude Code Is Steganographically Marking Requests
A blog post reveals that Claude models are embedding steganographic marks within their code responses. This practice could be used to trace the origin of AI-generated content or monitor model usage.
San Francisco's AI boom is pricing out six-figure tech workers
The AI boom in San Francisco is driving up the cost of living to such an extent that even couples earning $365,000 annually struggle to find affordable housing. Median rent is $3,827, and homes average $1.7 million, exacerbating housing affordability issues.
AI backfired – so Ford had to rehire humans
Ford discovered that hundreds of AI cameras used for design and manufacturing checks were prone to errors and pitfalls. This led the company to rehire experienced human engineers, dubbed "greybeards," to oversee these critical processes.
Libby will filter out AI content, kind of
OverDrive, the company behind the ebook lending app Libby, plans to implement a filter for AI-generated content. This move addresses concerns within the digital publishing industry about the influx of AI-produced books.
Start building with Nano Banana 2 Lite and Gemini Omni Flash
Google DeepMind announced the availability of Nano Banana 2 Lite and Gemini Omni Flash for developers. These new tools are designed to facilitate building with advanced AI models.
Fine-tune Amazon Nova models for accurate email data extraction
AWS is offering guidance on fine-tuning Amazon Nova models using Amazon SageMaker AI to improve email data extraction accuracy. This process can achieve up to 94.77% accuracy while reducing costs by 50%.
X now offers an MCP server to make its platform easier for AI tools to use
X has launched a hosted MCP server to simplify the integration of AI applications with its API. This aims to make the platform more accessible and functional for AI developers.
County with 37 Data Centers Asks Schools to 'Conserve Electricity'
Henrico County, Virginia, which hosts 37 data centers, has requested that local schools conserve electricity. This highlights the significant power demands placed on local grids by the proliferation of data centers, many of which support AI operations.
Microsoft takes down over 100 malicious Edge extensions hiding malware
Microsoft has removed 119 malicious Edge extensions that were found to be hiding malware within images and fonts. These extensions had been downloaded 2.6 million times since 2021.
UK watchdog plans to break Apple and Google’s ‘effective duopoly’ on mobile app stores
The UK's Competition and Markets Authority (CMA) is challenging Apple and Google's control over mobile app stores. The CMA aims to allow developers to direct users to alternative payment methods outside the app stores to foster greater competition.
AR Glasses Have a Massive Size Problem
According to Snap CEO Evan Spiegel, augmented reality (AR) glasses currently face a significant challenge with their physical size. This issue impacts user comfort and widespread adoption, as demonstrated by Snap's new Specs.