Rumors point to Claude 4.5 this week; OpenAI begins hiring for hardware
Get bigger weekly updates! Free subscribers receive the top stories each week, while Paid subscribers will get a few extra stories. All support for Handy AI directly helps me maintain the newsletter and keep the information flowing.
last week’s top stories
🔮 Rumors swirl of a Claude 4.5 launch this week. The AI community is abuzz that Anthropic might unveil a “Claude 4.5” model imminently, sparked by a credible leaker who correctly predicted Claude 4’s release. Developers have spotted hints on X and Reddit, fueling speculation that Claude’s next version could drop any day now. If true, the upgrade would mark the first major Claude boost in months, with fans hoping for improvements in reasoning, context length, and coding abilities. Read more
🤖 OpenAI poaches Apple talent for secret AI hardware. OpenAI, in partnership with ex-Apple design guru Jony Ive, has hired over two dozen Apple hardware engineers and designers as it plans to build AI-powered consumer devices. According to reports, OpenAI struck a deal with Apple’s Chinese supplier Luxshare to manufacture at least one new gadget, and is courting other Apple suppliers for parts. Read more
⚡ Elon Musk’s xAI launches Grok 4 Fast model. Musk’s new AI startup xAI introduced Grok 4 Fast, a cost-efficient reasoning model with a massive 2 million token context window. Grok 4 Fast is immediately available via the web, mobile apps and API, offering a unified model that can switch between “reasoning” and normal modes under one set of weights. It natively supports web browsing, X (Twitter) searches, and even handles images. Read more
📝 Notion rolls out AI “agent” to automate workspace tasks. Notion 3.0 introduced its first AI Agent, which can autonomously handle multi-step tasks across your notes, databases, and docs. The Agent draws on all your content as context and can create or update pages for you; for example, generating meeting notes, building a project tracker from Slack messages, or drafting a competitor analysis report. Unlike the previous Notion AI that only answered questions, the new Agent can execute 20-minute workflows, chaining actions to serve as a true “digital assistant” inside Notion. Read more
🤝 Nvidia invests $5 billion in Intel to jointly build AI chips. Nvidia announced a $5 billion investment for a ~4% stake in Intel, forging an alliance to develop next-gen PC and data center chips together. The move makes Nvidia one of Intel’s largest shareholders and comes weeks after the US government took its own stake to rescue Intel. Read more
🎥 Luma unveils Ray3, a “reasoning” AI for video generation. Startup Luma AI launched Ray3, calling it the world’s first video model that can reason through complex scenes. Ray3 breaks a video prompt into a plan (concept art, camera moves, lighting, etc.), generates drafts, self-critiques the results, and iteratively refines the output – essentially acting like an automated movie director. It’s also the first model to natively output high-dynamic-range 16-bit video (EXR files), meeting Hollywood-quality standards for color and detail. Read more
🕶️ Meta’s new smart glasses get a neural AI co-pilot. Meta launched its Ray-Ban Display smart glasses with a built-in display and an AI assistant on board. The glasses feature a tiny screen in one lens for notifications and real-time info, and come paired with a Meta Neural Band wristband that reads electrical signals from your arm, allowing you to control the glasses with subtle finger gestures. CEO Mark Zuckerberg touted the device as a step toward “personal superintelligence,” since it lets you ask Meta’s AI anything on the go (from translations in your view to reminders) without pulling out a phone. Read more
🧒 OpenAI adds teen safeguards and parental controls to ChatGPT. Facing pressure to protect younger users, OpenAI is rolling out an under-18 mode for ChatGPT with stricter content filters and new parental oversight tools. The AI will now refuse to engage in flirtation or self-harm discussions with teens, and if a minor user shows signs of severe distress, ChatGPT may alert parents or even authorities as a last resort. OpenAI is also building an age verification system and will “take the safer route” by defaulting uncertain users to the teen-safe experience until proven adult. These moves come as regulators and even 44 state attorneys-general scrutinize AI’s impact on kids, and OpenAI’s rivals like Meta are racing to implement similar teen protections. Read more
🏆 DeepMind AI beats almost everyone at coding competition. Google DeepMind is hailing a “historic” achievement after its latest AI model solved tough coding and math problems at a level matching the world’s top competitive programmers. In a global coding contest, the model (not officially named but hinted as part of Gemini) placed 2nd out of 139 champion human coders, even earning a gold medal ranking as if it were a top-20 coder worldwide. The AI had to tackle algorithmic problems requiring creativity and multi-step reasoning, not just rote code; one Google VP compared the breakthrough to Deep Blue’s win in chess, but for more real-world tasks. Read more
🧪 AI Research of the Week
Generative design of novel bacteriophages with genome language models
From Arc Institute & Stanford
Jake's Take: The team trained a genome language model (Evo 1/2) on millions of phage sequences, asked it to write whole viral genomes, then synthesized the designs in the lab.
About 300 genomes were built; 16 produced live phages that infected E. coli. The step-change is end-to-end genome design, moving beyond single-gene tinkering, which pushes phage therapy for drug-resistant infections closer to use. Powerful work that delivers real capability and demands clear biosecurity guardrails as these models scale.
and then, even more news…
🌐 Google embeds its Gemini AI assistant into Chrome. Google launched Gemini in Chrome for U.S. desktop users, turning the browser into an AI-enhanced helper. The built-in assistant can answer questions about your open tabs, summarize web pages or YouTube videos, and even help coordinate your schedule by integrating with Google Docs and Calendar. Chrome’s address bar is also getting an “AI mode” for smarter autocomplete and search suggestions, all part of Google’s push to make browsing more interactive and context-aware. Read more