Today:

• ChatGPT Hits 400M WAU, GPT-5 Nears

• Figure AI Unveils Helix

• AI Engineer Accelerates CUDA Kernels

• Spotify Expands AI Audiobook Narration

• Together AI Hits $3.3B Valuation

ChatGPT Hits 400M WAU, GPT-5 Nears

OpenAI’s ChatGPT has reached 400 million weekly active users, doubling enterprise adoption to 2 million businesses. Despite rising competition from xAI and DeepSeek, OpenAI is expanding into government contracts and Japan. GPT-5 is on the way, aiming to unify AI capabilities. Meanwhile, Musk’s takeover bid was rejected, and AI market battles intensify. The next year will determine whether OpenAI maintains its lead or faces disruption.

Figure AI Unveils Helix

Figure has launched Helix, a new robotics model that unifies vision, language, and action. It can handle entire humanoid upper-body control, pick up unknown objects, and collaborate with other robots on tasks. Through a two-part system, Helix processes high-level understanding while quickly adjusting actions in real-time. Trained on 500 hours of data, it needs no task-specific fine-tuning. Helix aims to accelerate household automation tasks by translating language instructions into precise robot movements.

AI Engineer Accelerates CUDA Kernels

The AI CUDA Engineer is an autonomous system that optimizes CUDA kernels, achieving 10–100x speedups in PyTorch operations. Developed by Sakana AI, it automates AI efficiency by converting, evolving, and refining CUDA code using AI-driven evolutionary techniques. The system outperforms existing CUDA implementations, enhancing AI model performance while reducing energy costs. Sakana AI envisions AI optimizing itself for faster, more sustainable computing, pushing the boundaries of efficiency in machine learning.

Spotify Expands AI Audiobook Narration

Spotify is expanding AI-narrated audiobooks through a partnership with ElevenLabs, allowing authors to use synthetic voices in 29 languages. While human narration remains a priority, Spotify sees AI as a cost-effective option for smaller authors and older books. AI-narrated titles will be clearly labeled. ElevenLabs' free version offers 10 minutes of narration per month, with premium plans reaching 500 minutes. This move could significantly increase AI-voiced audiobooks on the platform.

Together AI Hits $3.3B Valuation

Together AI has raised $305 million in a funding round led by General Catalyst, doubling its valuation to $3.3 billion. The Nvidia-backed startup specializes in cloud-based AI model training and deployment. Investors include Salesforce Ventures, Coatue, and Aramco’s Prosperity7. Together AI plans to expand its platform with Nvidia Blackwell GPUs and supports over 200 open-source AI models, positioning itself as a key player in the AI infrastructure market.

🧠RESEARCH

SuperGPQA: Scaling LLM Evaluation across 285 Graduate Disciplines

SuperGPQA is a new benchmark assessing large language models (LLMs) across 285 graduate disciplines, beyond traditional fields like math and science. Using expert feedback and AI collaboration, it refines evaluation questions. Findings show LLMs struggle in many areas, with top models achieving only 61.82% accuracy, highlighting gaps in AI capabilities.

SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features

SigLIP 2 enhances multilingual vision-language models with improved semantic understanding, localization, and dense features. It integrates captioning-based pretraining, self-supervised learning, and online data curation. Outperforming its predecessor, it excels in classification, retrieval, and visual tasks. Trained on diverse, de-biased data, it improves fairness and multilingual capabilities across four model sizes.

S*: Test Time Scaling for Code Generation

S* introduces a test-time scaling framework for code generation, enhancing model accuracy through hybrid parallel and sequential scaling. It improves solution selection using adaptive input comparisons and execution-based validation. S* boosts performance across 12 models, enabling smaller models to outperform larger ones, with significant gains on LiveCodeBench benchmarks.

LongWriter-V: Enabling Ultra-Long and High-Fidelity Generation in Vision-Language Models

LongWriter-V enhances vision-language models (VLMs) for ultra-long, high-fidelity text generation. It introduces a 22k-example dataset and an iterative correction method (IterDPO) to improve coherence. The model excels in generating up to 10,000-word outputs, surpassing larger proprietary models like GPT-4o on long-form writing benchmarks.

PC-Agent: A Hierarchical Multi-Agent Collaboration Framework for Complex Task Automation on PC

PC-Agent is a hierarchical multi-agent system designed to automate complex PC tasks. It improves perception with an Active Perception Module and enhances decision-making through instruction decomposition, progress tracking, and adaptive error correction. PC-Agent outperforms previous models, achieving a 32% higher task success rate on real-world PC automation benchmarks.

🛠️TOP TOOLS

LogoAI - Creating professional logos, cohesive visual identities, and automating brand promotion.

BoredHumans - Offers over 100 free AI tools, designed to cater to a wide range of creative, entertainment, and productivity needs.

Duplichecker - Designed to ensure the originality and integrity of written content.

Mnml AI - AI-powered design assistant specifically designed for architects, interior designers, and related professionals.

Getimg - AI-driven image generation and editing tool designed to create high-quality visuals from text descriptions.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Uber uses AI to enhance customer support, optimize rides and deliveries, and improve workforce productivity. AI-driven personalization, automation, and efficiency help Uber scale operations while driving innovation in the on-demand economy.

• Nvidia has launched "Signs," an AI-powered platform that helps users learn American Sign Language through a 3D avatar and real-time feedback. The free tool aims to expand accessibility while showcasing Nvidia’s broader AI ambitions.

• AI hiring startup Mercor secured $100 million in Series B funding, reaching a $2 billion valuation. The company uses AI to streamline hiring, boasting rapid revenue growth and plans for broader applications in education and workforce automation.

• Google's AI tool solved a decade-long superbug mystery in just two days, replicating unpublished research and offering new insights. Scientists believe this breakthrough could revolutionize scientific discovery and accelerate medical advancements.

• Researchers developed an AI tool that diagnoses diseases like diabetes, HIV, and COVID-19 from a single blood sample. By analyzing immune-cell gene sequences, this breakthrough could revolutionize diagnostics for conditions with overlapping symptoms.

• Researchers trained AI models on novels from different decades to analyze historical social biases. By examining chatbot responses, they track how racism and sexism have evolved over time, offering a new way to study cultural attitudes.