A New Chapter in Home Construction

What if new-home construction were nearly as easy as opening a book? That’s the story with BOXABL.

How? By using assembly lines to condense homebuilding from 7+ months to hours, BOXABL ships readymade houses to their final destination. Then, they’re unfolded and immediately livable.

They’ve already built 700+. But the real transformation’s still coming.

BOXABL’s currently preparing for Phase 2 – combining modules into larger townhomes, single-family homes, and apartments. And until 6/24, you can join as an investor for just $0.80/share.

They already fully maxed out a $75M investment campaign once, so don’t wait around.

Invest in BOXABL Now

_{*This is a paid advertisement for Boxabl’s Regulation A offering. Please read the offering circular at https://invest.boxabl.com/#circular}

Today:

• o3-Pro Excels At Deep Reasoning

• Mistral Launches Magistral Reasoning Model

• Zuckerberg Builds Meta Superintelligence Team

• OpenAI Strikes Google Cloud Deal

• OpenAI Slashes o3 Model Pricing

o3 pro is a BEAST... one-shots Apple's "Illusion of Thinking" test

OpenAI’s new o3-pro model trades speed for deep, reliable thinking, solving complex puzzles like a 10-disk Tower of Hanoi that stumped other AIs. It digests large context, writes multi-agent code scaffolds, and generates company roadmaps once given extensive data. 

Early users must treat it as a slow report generator, not a chat bot. Meanwhile, original o3 becomes 80 % cheaper, widening access while benchmarks struggle to capture o3-pro’s real-world jump gains.

WATCH THE VIDEO ON YOUTUBE

Mistral Launches Magistral Reasoning Model

Mistral AI unveils Magistral, its first reasoning model, offered as open-source 24-billion-parameter (size measure) Magistral Small and business-ready Magistral Medium. The model solves subject-focused problems, showing clear step-by-step thinking in many languages, and scores strongly on AIME 2024. New Think mode and Flash Answers yield ten-times faster replies in Le Chat. Magistral helps law, finance, coding, and storytelling, inviting community upgrades through an open license release that starts today worldwide.

Why this matters

• See-through thinking – Magistral prints each reasoning step, so anyone can check and fix its logic, boosting trust in AI answers.
  

• Shared freely – The small version is open-source (publicly shared), letting researchers and builders study and improve a strong reasoning model without big fees, speeding progress.
  

• Talks many languages fast – Multilingual skill plus 10× quicker replies bring deep reasoning to users everywhere and many fields, from law to code.

Zuckerberg Builds Meta Superintelligence Team

Mark Zuckerberg is personally assembling a “superintelligence” unit inside Meta after feeling the firm lags in artificial-intelligence breakthroughs. He has invited elite researchers and engineers to private meetings in Lake Tahoe and Palo Alto, asking them to join a secret team tasked with reaching artificial general intelligence—machines that can match humans at many jobs. Zuckerberg wants Meta to beat every rival in this race and is giving it top priority.

Why this matters

• A tech-giant CEO’s hands-on push signals an even fiercer global race to master human-level AI.
  

• Chasing artificial general intelligence forces bigger research budgets and faster progress across the whole field.
  

• Poaching top scientists reshapes talent pools, partnerships, and the pace at which new ideas reach everyday products.

OpenAI Strikes Google Cloud Deal

OpenAI will use Google Cloud's computing power under a deal finalized in May, easing its reliance on Microsoft Azure and meeting rising demand to train and run its language models. The move adds a rival to Google’s customer list, boosting Google Cloud while forcing it to juggle its own AI needs. Alphabet shares rose on the news, showing investors' faith in the cloud unit’s growth despite ChatGPT challenging Google search.

Why this matters

1. Diversifies OpenAI’s supply – Using two big clouds lowers the risk of shortages and price spikes for the computing power that drives new AI models.
   

2. Strengthens Google Cloud – Landing a marquee customer proves Google’s chips and data centers can compete with Microsoft and Amazon, attracting more AI startups.
   

3. Redraws rivalry lines – Competitors are now also partners, hinting that future breakthroughs may depend on cooperation as much as head-to-head races.

🧠RESEARCH

Saffron-1: Towards an Inference Scaling Paradigm for LLM Safety Assurance

SAFFRON-1 is a new method to improve AI safety during use, not just during training. It uses a more efficient way to check if AI responses are safe, cutting down on expensive safety checks. This makes AI harder to break and more reliable under real-world threats.

MiniCPM4: Ultra-Efficient LLMs on End Devices

MiniCPM4 is an efficient language model built for devices like phones or laptops. Using smarter attention methods, cleaner data, and faster training techniques, it delivers strong performance while using fewer resources. MiniCPM4 runs faster than similar models, handles long text better, and can power real-world tasks directly on user devices.

OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation

OneIG-Bench is a new benchmark to better test AI models that turn text into images. It checks not just image quality, but also how well models understand reasoning, render text, handle styles, and create diverse outputs. This helps identify strengths and weaknesses, guiding improvements in future image generation systems.

🛠️TOP TOOLS

PromptPal - AI prompt management platform designed to streamline the process of creating, organizing, and collaborating on AI-generated content.

Tldv - AI-powered meeting assistant designed to streamline the entire meeting process, from recording to follow-up.

GPT Researcher - open-source autonomous agent designed to revolutionize online research tasks.

Designify - AI-powered web application that transforms ordinary photos into professional-quality designs.

Playground AI - Text-to-image model developed by Playground AI

📲SOCIAL MEDIA

🗞️MORE NEWS

• OpenAI cut o3 model prices by 80%, making advanced AI much cheaper for developers. The move pressures rivals like Google, Anthropic, and DeepSeek, while opening access for startups and smaller teams to experiment and build affordably.

• Google revamped its internal learning platform, Grow, to focus on AI training for employees. The shift prioritizes business needs, cuts unused courses, and reflects Google’s intense focus on AI amid cost-cutting and competitive pressure.

• Snap will launch AI-powered AR glasses called Specs in 2026. The glasses combine AI assistants with 3D experiences, aiming to move beyond smartphones. Snap sees wearables and spatial computing as key to its future growth.

• Enterprise AI startup Glean raised $150 million at a $7.2 billion valuation, tripling its value since early 2024. Its AI-powered search and agent tools help employees access corporate data across platforms like Salesforce and Teams.