Read Online | Sign Up | Advertise | Upgrade
Presented by
Good morning, AI enthusiasts! President Trump just launched the Genesis Mission, a federal AI initiative he's comparing to the Manhattan Project, and with plans to integrate decades of government datasets to train scientific foundation models, the comparison might not be hyperbole.
With 270 days to demonstrate working capability and $30B compute deals from OpenAI and Anthropic fresh in memory, America's research infrastructure is shifting from human-paced to AI-accelerated.
In today's recap:
White House commits federal datasets to AI training
Claude Opus 4.5 beats human coders on tests
New AI tools & prompts
WHITE HOUSE
Trump launches AI Manhattan Project
The White House
Recaply: President Trump signed an executive order Monday launching the Genesis Mission, a federal AI initiative compared to the Manhattan Project that will integrate government datasets to train scientific foundation models, according to the White House.
Key details:
The American Science and Security Platform will provide researchers with DOE supercomputers, AI agents for automated experimentation, and secure access to federal scientific datasets spanning decades of government investment.
Secretary of Energy Chris Wright must identify computing resources within 90 days, compile initial datasets within 120 days, and demonstrate working capability for at least one national challenge within 270 days.
The initiative targets 20 science challenges spanning advanced manufacturing, biotechnology, critical materials, nuclear energy, quantum computing, and semiconductors, with annual updates reflecting progress.
Michael Kratsios will lead coordination across agencies through the National Science and Technology Council, with plans for competitive research fellowships at national laboratories.
Why it matters: OpenAI's $30B AWS deal and Anthropic's $30B Azure commitment signal the compute arms race is accelerating. Now the federal government joins with the world's largest scientific datasets and laboratory infrastructure. If executed well, this shifts America's research from human-paced to AI-accelerated.
PRESENTED BY ROKU
Shoppers are adding to cart for the holidays
Peak streaming time continues after Black Friday on Roku, with the weekend after Thanksgiving and the weeks leading up to Christmas seeing record hours of viewing. Roku Ads Manager makes it simple to launch last-minute campaigns targeting viewers who are ready to shop during the holidays. Use first-party audience insights, segment by demographics, and advertise next to the premium ad-supported content your customers are streaming this holiday season.
Read the guide to get your CTV campaign live in time for the holiday rush.
ANTHROPIC
Claude Opus 4.5 beats human coders
Anthropic
Recaply: Anthropic just released Claude Opus 4.5, featuring a new effort parameter that lets developers control reasoning depth, achieving state-of-the-art results on software engineering benchmarks with dramatically fewer tokens.
Key details:
The effort control lets developers choose speed or capability, with medium effort matching Sonnet 4.5 while using 76% fewer output tokens, and high effort exceeding Sonnet by 4.3 points using 48% fewer tokens.
Opus 4.5 scored higher than any human candidate on Anthropic's performance engineering exam, completing complex debugging and system design within the 2-hour limit.
The model found creative workarounds on tau-bench by upgrading cabin class before modifying flights, a legitimate solution the benchmark scored as failure but represents advanced reasoning.
Claude Code gains Plan Mode building editable plan files before execution, now available in desktop apps for parallel sessions alongside expanded Excel beta access.
Why it matters: A dial for AI reasoning felt inevitable, and Anthropic shipped it first at scale. While other models overthink simple tasks or underperform on complex ones, Opus 4.5's effort parameter gives developers control over speed versus capability tradeoffs.
NEWS
What Matters in AI Right Now?
OpenAI rolled out Shopping Research feature in ChatGPT enabling product discovery and purchase recommendations through conversational AI interactions across merchants
Zyphra announced it has successfully trained ZAYA-1, the first large-scale Mixture-of-Experts model entirely on AMD Instinct MI300X GPUs, outperforming Llama 3 8B across multiple benchmarks
Hapiko launched Stickerbox, a $99 voice-powered AI printer turning kids' spoken ideas into physical stickers within seconds, backed by $7M funding
AWS announced up to $50B investment to expand AI and supercomputing infrastructure for US government agencies, adding 1.3 gigawatts of capacity by 2026
Google in talks to sell Tensor Processing Unit chips to Meta, potentially challenging Nvidia's AI chip dominance with custom TPU infrastructure deals
UAE announces $1B AI initiative for African infrastructure development, partnering with Kenya and Ghana to establish data centers and innovation hubs across continent
Tencent released HunyuanOCR as open source, expanding its AI document processing capabilities for text extraction and recognition across multiple languages
TOOLS
AI Tools to Check Out
📚 MyReader: Chat, summarize, or turn docs into audiobooks
🗒️ Audionotes: Turn voice recordings into organized text notes
🌐 Rask: Dub and translate videos with AI voice cloning
🔎 Coral AI: Ask questions about documents with cited answers
🤖 Dante: Build custom chatbots on your business data
🧭 HIX: All‑in‑one writer, chatbot, and browser assistant
🎬 Pollo: Create videos from text prompts and images
🧠 Cody: AI assistant trained on your company knowledge
* Some links in this newsletter may be from sponsors or affiliates. We may get paid if you buy something through these links.
PROMPTS
Draft Employee Survey Questions
Write 6–8 employee survey questions designed to measure [e.g. belonging, manager trust, workload balance]. Ensure the questions are neutral and easy to understand. Format them as one question per line with rating scale suggestions.EVENTS
Lovable SheBuilds Season 2 - Dec 15-17, 2025
ElevenLabs Worldwide Hackathon - Dec 11, 2025
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
