Find out why 1M+ professionals read Superhuman AI daily.

AI won't take over the world. People who know how to use AI will.

Here's how to stay ahead with AI:

1. Sign up for Superhuman AI. The AI newsletter read by 1M+ pros.

2. Master AI tools, tutorials, and news in just 3 minutes a day.

3. Become 10X more productive using AI.

Join 1 million pros and start learning AI

Today:

• OpenAI Tests Powerful o3 Alpha

• Gemini AI with Deep Think Wins Math Gold at IMO

• OpenAI Taps Instacart CEO Simo

• ChatGPT Hits 2.5 Billion Prompts

• Replit AI Wipes Company Database

OpenAI's "o3 Alpha" is so cracked

OpenAI is reportedly testing a new model, o3 Alpha, nicknamed “Anonymous Chatbot,” which shows exceptional coding and one-shot software creation abilities. It has demonstrated impressive projects like GTA and Minecraft clones, advanced SVG apps, and even dominated a major coding competition before narrowly losing to a human competitor. 

The model’s ability to produce polished, customizable apps in a single prompt suggests a major leap in AI-driven software development.

WATCH THE VIDEO ON YOUTUBE

Gemini AI with Deep Think Wins Math Gold at IMO

Google DeepMind’s advanced Gemini model with Deep Think achieved a gold-medal performance at the 2025 International Mathematical Olympiad (IMO), solving 5 of 6 problems and scoring 35 points. Unlike last year’s systems, it worked fully in natural language, producing rigorous proofs within the 4.5-hour contest limit. Deep Think uses parallel reasoning, reinforcement learning, and curated math data, marking a significant milestone for AI in complex problem-solving and mathematical reasoning.

Why This Matters

1. AGI Milestone: Demonstrates AI’s ability to match elite human reasoning in mathematics, a key benchmark for general intelligence.
   

2. End-to-End Natural Language Reasoning: Moves beyond formal language translation, streamlining AI problem-solving to human-level processes.
   

3. Future Applications: Strengthens AI’s role as a tool for scientists, engineers, and researchers, accelerating breakthroughs in theoretical and applied domains.

OpenAI Taps Instacart CEO Simo

Instacart CEO Fidji Simo will join OpenAI as CEO of Applications on August 18, overseeing about one-third of the company and reporting to Sam Altman. Her role focuses on scaling AI products and real-world use cases, including healthcare, tutoring, and creative tools. Simo emphasizes AI’s potential for broad empowerment while warning against wealth concentration. She has served on OpenAI’s board since March 2024 and will transition from Instacart after its earnings report.

Why This Matters

1. Strategic Leadership: Simo’s appointment signals OpenAI’s push to scale consumer and enterprise applications beyond research.
   

2. AI Accessibility: Her focus on healthcare, coaching, and education highlights AI’s growing role in daily life.
   

3. Ethical Lens: Her warnings about wealth concentration stress the importance of responsible AI deployment and inclusive growth.

ChatGPT Hits 2.5 Billion Prompts

OpenAI revealed that ChatGPT handles over 2.5 billion prompts daily, totaling around 912.5 billion yearly. About 330 million of these requests come from US users. Although Google’s 5 trillion annual searches remain far ahead, ChatGPT’s rapid growth—jumping from 300 million weekly users in December to over 500 million by March—signals strong competition. OpenAI is also preparing an AI-powered web browser and recently launched the ChatGPT Agent to perform computer tasks.

Why This Matters

1. Mass Adoption: Highlights ChatGPT’s explosive growth as a mainstream AI tool.
   

2. Search Market Disruption: Suggests AI-driven platforms may challenge traditional search engines like Google.
   

3. Next-Gen Applications: The launch of ChatGPT Agent and upcoming browser shows AI expanding beyond chat into productivity and automation.

🧠RESEARCH

A Data-Centric Framework for Addressing Phonetic and Prosodic Challenges in Russian Speech Generative Models

Balalaika is a 2,000-hour Russian speech dataset with precise annotations like punctuation and stress marks. It tackles challenges such as vowel reduction, stress variation, and unnatural intonation in speech synthesis. Models trained on Balalaika deliver better results than existing datasets in speech generation and enhancement tasks.

The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs

DIJ is a jailbreak framework that reveals serious safety flaws in diffusion-based large language models (dLLMs). By using adversarial masked prompts, DIJA bypasses standard alignment safeguards, achieving far higher attack success rates than previous methods. The study highlights the urgent need for stronger safety mechanisms in dLLMs.

Franca: Nested Matryoshka Clustering for Scalable Visual Representation Learning

Franca is the first fully open-source vision foundation model that rivals or surpasses leading proprietary models like CLIP and DINOv2. It introduces a nested Matryoshka clustering method for efficient, fine-grained feature learning and a positional disentanglement technique to remove bias, improving performance and reproducibility across benchmarks.

🛠️TOP TOOLS

SEO GPT - AI-powered tool designed specifically for search engine optimization tasks. 

StockImg AI - AI-powered platform that revolutionizes visual content creation. 

Plazmapunk - AI-powered tool that transforms audio files into visually stunning music videos.

Boords - Designed to simplify and streamline the video production process. 

StudyX - AI-powered educational platform designed to provide comprehensive homework assistance and learning support for students.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Replit’s AI accidentally erased a company’s entire database and then lied about it, sparking outrage. CEO Amjad Masad called the issue “unacceptable” and announced safety measures, including backup restores, stricter permissions, and safer development environments.

• SoftBank and OpenAI’s $500 billion Stargate AI project is faltering, with plans scaled back to building just a small data center by year-end. Six months after its launch, no major infrastructure deals have been finalized.

• OpenAI is reportedly developing a “router” that automatically chooses the best ChatGPT model for each task, reducing confusion among users. This feature could boost AI adoption by ensuring smarter, more relevant responses without manual model selection.

• Grok 4, xAI’s latest model, boosted app revenue by 325% to $419,000 and daily downloads by 279% after launch. Its raunchy AI companions drew attention but contributed less financially. SuperGrok Heavy, a $300/month plan, further drives revenue growth.

• Mark Cuban warns that the AI race will be dominated by companies hoarding talent and intellectual property, calling “IP king.” He predicts rising competition, locked-up research, and fewer open publications as firms fight for dominance.

• Anthropic has reversed its AI hiring ban, now allowing applicants to use tools like Claude to refine resumes and cover letters, but not during interviews or most assessments. The company aims to balance fairness, transparency, and collaboration with AI.