- Recaply AI
- Posts
- Trump drops AI dominance blueprint
Trump drops AI dominance blueprint
+ AI solves 2,000-year Roman mysteries
July 24, 2025 |
Good morning, AI enthusiasts. The Trump administration just released America's AI Action Plan, outlining how the U.S. will maintain global AI leadership while blocking adversaries from accessing American technology.
With this blueprint focusing heavily on export controls and international alignment, are we witnessing the formalization of a new AI cold war?
Friendly reminder: Our newsletter is still slowly moving to a new email address. Add [email protected] to your contacts to ensure you keep receiving our emails.
In today’s recap:
White House unveils comprehensive AI dominance strategy
Google's Aeneas AI decodes 2,000-year-old Roman mysteries
Kafka AI employee autonomously completes real freelance work
Brain-inspired AI outsmarts massive models with 1000x fewer resources
5 new AI tools & prompts
LATEST DEVELOPMENTS
THE WHITE HOUSE

The White House / Screenshot
Recaply: The Trump administration just released America's AI Action Plan, outlining a strategy to maintain global AI leadership while blocking adversaries from accessing U.S. technology.
Key details:
Plan focuses on strengthening AI chip export controls through "creative approaches" and location verification features.
Calls for allies to follow U.S. export restrictions rather than "backfill" to adversaries.
Administration may use Foreign Direct Product Rule and secondary tariffs for international alignment.
Document remains light on specific implementation details, focusing on foundational frameworks.
More detailed actions may come in the form of executive orders expected on next few day.
Why it matters: This plan shows the tough balance the U.S. is trying to manage growing the AI market worldwide while still leading in the field. Some mixed messages, like undoing rules from the Biden era, show that the U.S. is still figuring out its best strategy in a fast-changing global AI race.
GOOGLE DEEPMIND

Google DeepMind
Recaply: Google DeepMind just released Aeneas, a breakthrough AI model that helps historians analyze ancient Roman inscriptions by finding connections between fragmentary texts that have puzzled researchers for decades.
Key details:
Aeneas looks at over 176,000 Latin inscriptions to find “parallels” texts that are similar and give helpful context.
Uses both text and images to restore damaged inscriptions, predict dates, and identify locations.
Achieves 73% accuracy in restoring missing text gaps and 72% accuracy in geographic attribution.
In tests, historians said Aeneas gave helpful results 90% of the time, which made them feel more confident in their work.
Available free at predictingthepast.com with open-source code and educational materials.
Why it matters: Old Roman writings are like puzzle pieces many are broken or missing. Aeneas gives historians a new way to connect these pieces and learn more about everyday life in ancient Rome, helping uncover stories that have been hidden for hundreds of years.
BRAINBASE LABS

Brainbase Labs
Recaply: Brainbase Labs introduced Kafka, an AI employee that demonstrated versatility by creating commercials, deploying cloud infrastructure, and completing freelance tasks autonomously.
Key details:
Created and produced a commercial using Google Veo 3 for its own product launch.
Deployed production microservices on AWS Lambda and EventBridge from scratch.
Successfully completed various Upwork and Fiverr tasks across different skill areas.
Achieved 77.2% accuracy on GAIA Level 3 benchmark for economically viable knowledge work.
Operates through email, Slack, and phone like a standard remote employee.
Why it matters: Kafka proves that AI is no longer just a cool idea it can actually do useful jobs in the real world. From creative work to coding, it shows how AI can help human teams in many different industries.
AI BREAKTHROUGHS
SAPIENT INTELLIGENCE

Sapient Intelligence
Recaply: Singapore-based Sapient Intelligence just released the Hierarchical Reasoning Model (HRM), a new AI system built to think more like the human brain and has shown amazing results, even beating much bigger AI models.
Key details:
HRM uses just 27 million parameters and 1000 training examples to solve complex puzzles that stump much larger AI models.
Scored 40.3% on the ARC-AGI test—better than OpenAI’s o3-mini-high (34.5%) and Claude 3.7 (21.2%).
Solved 55% of extreme Sudoku puzzles and 74.5% of hard maze problems while other top models scored 0%.
Features two connected modules: high-level for strategic planning, low-level for detailed calculations.
Works without pre-training or chain-of-thought prompting that current AI systems rely on.
Why it matters: This brain-like design may be a big change in how AI is built. Instead of needing huge resources, HRM shows that smart design can do more with less. It could help make powerful AI tools easier to build and use.
FAST TRACKS
📰 What Matters in AI Right Now?
Microsoft launched GitHub Spark in public preview, enabling users to build full-stack intelligent apps using natural language with Claude Sonnet 4 integration.
Lovable introduced Lovable Agent with fully agentic capabilities, reducing errors by 91% and enabling complex multi-step development tasks.
Google reported AI Overviews reached 2 billion monthly users, up from 1.5 billion in May 2025 across 200 countries and territories.
Meta hired three Google DeepMind researchers who contributed to the company's International Mathematics Olympiad gold medal-winning model.
Perplexity added MCP support to its macOS app, introducing a new Connectors option in settings for enhanced functionality.
Firecrawl launched Firecrawl Observer, an open source web monitor for tracking page changes with custom intervals and webhook alerts.
🧰 AI Tools of the Day
📢 Quickads: Generate scroll-stopping image & video ads fast with AI.
🗣️ Otter: Transcribe, organize, and summarize meetings in real time.
📝 Circleback: Capture and automate meeting notes across email and calendars.
📄 JobWinner: Create tailored resumes and cover letters using AI.
🎬 Videotowords: Transcribe videos/audio into text with multi-language support.
AI PROMPTS
📘 Python Learning Plan Generator
#CONTEXT:
Adopt the role of an expert Python instructor with extensive experience in teaching programming concepts to beginners. Your task is to create a comprehensive Python learning plan that will guide a novice programmer from the basics to more advanced topics. The plan should be structured, engaging, and designed to foster a deep understanding of Python programming.Important Note: Click the button below to see the full prompt in Notion
Have a favorite prompt? Tell us about it or rate today’s prompt here
That’s a wrap for today!
We hope today’s insights saved you time and brought more clarity to your day.
👉️ Got a second, Your feedback help us make Recaply even better.
How did today’s issue land for you?Your feedback helps me create better emails for you! |
Have an idea for Recaply AI?
We’re all ears. Send your thoughts to [email protected] — we read every message and love turning your ideas into reality.
See you in the next recap,
— Jason @ Recaply AI

Reply