The new hub for Open AI

+ Japan uses AI to simulate Mount Fuji eruption

August 28, 2025

Good morning, AI enthusiasts. Prime Intellect just opened its AI Training Hub to all researchers, giving everyone free access to the same advanced training tools that big tech companies use.

How might this democratization of AI development reshape the competitive landscape?

In today's recap:

  • Prime Intellect opens AI training to all

  • OpenAI & Anthropic collaborate on safety

  • Japan simulates Mount Fuji eruption with AI

  • OpenAI enhances Codex for developers

  • WhatsApp adds AI message writing assistant

PRIME INTELLECT

Prime Intellect opens AI Training Hub to all

Environments Hub

© Prime Intellect

Recaply: Prime Intellect just launched its Environments Hub, a community platform where researchers share AI training environments. The company opened this tool to everyone after a successful private beta with over 30 contributors. This platform aims to help open-source AI models compete with big tech companies.

Key notes:

  • Over 30 researchers and companies joined the private beta testing phase.

  • The hub provides free access to reinforcement learning training environments.

  • Prime Intellect offers bounties and funding for novel environment creators.

  • The entire technology stack remains completely open-source for public use.

  • Platform supports their upcoming INTELLECT-3 AI model development.

Impact: This launch could level the playing field between open-source AI developers and major tech companies, giving smaller teams access to advanced training tools previously available only to big labs.

Share on FacebookShare on XShare on LinkedIn
ANTHROPIC & OPENAI

AI safety gets boost from competitor collaboration

On the Password Protection evaluations set, Opus 4 and Sonnet 4 both matched OpenAI o3 in perfect 1.000 performance. This matches our previous findings that reasoning models do well on this task.

© OpenAI

Recaply: In a surprising move, OpenAI and Anthropic decided to test each other's AI models for safety problems. The companies looked for issues like fake information and harmful behavior patterns. Their joint effort represents a new approach to making AI safer through cooperation.

Key notes:

  • The evaluation covered multiple safety areas including sycophancy, whistleblowing, and deceptive behaviors.

  • Tests showed trade-offs between being helpful and being accurate across different AI models.

  • Both companies relaxed some security measures to allow thorough testing by their competitors.

  • Results revealed that no single model performed best in all safety categories.

  • The initiative aims to establish industry-wide standards for AI safety evaluation.

Impact: This collaboration proves that AI safety can bring together even competing companies. As AI becomes part of daily life for millions of people, having multiple companies work together on safety testing helps protect everyone from potential risks.

Share on FacebookShare on XShare on LinkedIn
JAPAN & AI

Japan uses AI video to show Mount Fuji eruption impact

A screenshot from an Al-generated video showing what could happen to Tokyo if Mount Fuji erupted

© Screenshot

Recaply: Japan's government released a dramatic AI-generated video showing what would happen if Mount Fuji erupted. The video demonstrates how volcanic ash could reach Tokyo within two hours, affecting 20 million residents. Officials created this realistic simulation to help people prepare for a potential disaster.

Key notes:

  • AI video shows ash clouds covering Tokyo's famous skyline and streets.

  • Mount Fuji last erupted 318 years ago but remains an active volcano.

  • Economic damage could reach $16.6 billion from a major eruption.

  • Even small amounts of ash could shut down trains and block roads.

  • Government recommends keeping two weeks of emergency supplies.

Impact: This innovative use of AI technology helps make disaster preparedness more real for Tokyo residents. While some critics call it alarmist, the visual approach could save lives by encouraging better preparation.

Share on FacebookShare on XShare on LinkedIn
NEWS

📰 What matters in AI right now?

  • OpenAI just upgraded Codex with IDE extension support, GitHub code reviews, cloud-to-local task handoff, and internet access powered by GPT-5.

  • Moonshot AI introduced Kimi PPT, a new feature to turn text-to-PPTs within minutes

  • WhatsApp introduced Writing Help, an AI assistant that suggests message rewrites in professional, funny, or supportive styles while keeping chats private.

  • Microsoft is set to bring Copilot to Samsung 2025 TVs and monitors for free, turning TVs into AI companions with voice commands and visual cards.

  • Tencent announced the open-source release of HunyuanVideo-Foley, generating 48kHz professional-grade audio effects for videos with state-of-the-art performance.

  • Google just upgraded Vids with AI avatars, image-to-video via Veo 3, automatic transcript trimming, and a free version, reaching 1M monthly users.

  • Liquid AI released LFM2-VL multimodal models (450M and 1.6B parameters) designed for 2x faster GPU inference on edge devices.

  • Nous Research introduced Hermes 4, hybrid reasoning models combining structured multi-step reasoning with instruction-following in 405B, 70B, and 14B sizes.

  • Anthropic detected and stopped cybercriminals using Claude Code for data theft and extortion, targeting 17+ organizations with ransoms over $500,000.

OUR TOP PICKS

🧰 AI Tools to Check Out

  • Make: Connect apps and automate workflows without coding.

  • 🎙️ Clipto: Convert audio and video to text with AI transcription.

  • 🚀 HighLevel: All-in-one marketing platform for agencies.

  • 📊 IQ Hashtags: Optimize Instagram reach with smart hashtag research.

  • 🏢 Hubflo: Create branded client portals for project collaboration.

  • 🎧 Podcastle: Create professional podcasts with AI editing tools.

  • 🤖 Monica: AI assistant browser extension with multiple models.

  • 💬 Tidio: Automate customer support with AI chatbots and live chat.

PROMPTS

📝 Optimize To-Do List System

#CONTEXT:
You are a productivity expert who has been tasked with helping a user develop an effective and optimized to-do list system. Your goal is to provide comprehensive guidelines and best practices to ensure maximum productivity and goal achievement.

#ROLE:
Adopt the role of a highly organized productivity expert with a deep understanding of effective task management systems.

Important Note: This is not the full prompt. You’ll need to click the button below to get the complete prompt.

Have a favorite prompt? Tell us about it or rate today’s prompt, click here.

🧡 Enjoyed this issue? 

🤝 Recommend our newsletter or leave a feedback.

How'd you like today's newsletter?

Your feedback helps me create better emails for you!

Login or Subscribe to participate in polls.

Cheers, Jason

Connect on LinkedIn, & Twitter.

Reply

or to participate.