Weekly AI #14

Welcome to Weekly AI #14. This week, advancements in humanoid robotics bring us closer to AGI, with Figure Helix and Neo Gamma showcasing unprecedented capabilities in reasoning, learning, and interaction.

Top 5 Tools

Kokoro TTS - A cutting-edge AI text-to-speech model with 82M parameters, built on StyleTTS 2 architecture

Best Free Tools - A database of the best free tools used by indie founders & creators

SEO AI Agent - Turn weeks of SEO work into minutes with the most advanced SEO AI Agent

Nim - All-in-one AI video app

ShakkerAI - If you can dream it,Shakker can stream it

Top News

Elon Musk’s xAI Launches Grok 3 - Elon Musk's xAI launches Grok 3, boasting capabilities in math, science, and coding, surpassing current AI models. The model includes enhanced reasoning and DeepSearch, utilizing 200,000 H100 GPUs. Accessible to subscribers, Grok 3 aims to set new AI standards amid rising competition.

Figure Robots Unveils Helix - Helix, the latest innovation by Figure Robots, revolutionizes robotics with its human-like reasoning and dexterity, enabling seamless adaptation to various tasks and environments. Powered by a unique AI architecture, Helix demonstrates remarkable generalization and collaboration, bringing us closer to everyday multifunctional robots.

Shorts

Mon, Feb 17
- Sam Altman says GPT-4.5 sparks "feel the AGI" moment among high-taste testers
- NVIDIA unveils Magic 1-For-1 tool to generate one-minute video clips in one minute
- South Korea suspends DeepSeek app downloads over personal data protection rule violations
- Mistral AI introduces Saba, a 24B model for Middle Eastern and South Asian languages
- Elon Musk’s xAI plans to launch an AI game studio

Tue, Feb 18
- Suno updates Edit feature with fresh UI Fade Out and streamlined tools for seamless track edits
- GitHub Copilot launches GPT-4o mini-based model with faster code completions in VS Code
- Google Gemini Deep Research now available globally on mobile and desktop for comprehensive reports
- Meta announces LlamaCon AI conference for April 29 2025 to highlight open-source AI advancements
- Mira Murati launches Thinking Machines Lab for customizable AI solutions
- Perplexity AI open-sources R1 1776 model for uncensored unbiased factual information delivery
- Elon Musk says Grok 3 to improve daily this week urges users to report issues for updates
- OpenAI launches SWE-Lancer benchmark for real-world AI coding performance evaluation
- Pika Labs launches official iOS app
- NBA unveils AI robots for rebounding coaching and player wellness at 2025 All-Star Tech Summit
- Kimi AI introduces MoBA for efficient long-context LLM processing with 6.5x speedup at 1M input
- DeepSeek introduces Native Sparse Attention for faster AI training and long-text efficiency
- Elon Musk says xAI's next training cluster to have 1M GPUs worth $25B-$30B for advanced AI workloads
- SkyReels-A1 enables expressive portrait animation in video diffusion transformers with facial expression-aware landmarks
- xAI announces Grok's upcoming speech capabilities for advanced AI interaction
- Elon Musk says SpaceX to send Starship rockets to Mars with Optimus robots and Grok in 3 years
- SkyReels V1 introduces human-centric video model with advanced facial animation and I2V tech
- xAI unveils Grok 3
- Sam Altman sparks debate on ChatGPT o3-mini vs phone-optimized AI model for broader accessibility
- Fal AI adds SkyReels V1 open-source human-centric video model
- Ilya Sutskever raises $1B+ for Safe Superintelligence startup at $30B+ valuation in 4 months

Wed, Feb 19
- Google AI releases PaliGemma 2 mix for image captioning OCR and object detection in various sizes
- Arc Institute and NVIDIA unveil Evo-2 AI model trained on 9.3T DNA base pairs for genome writing
- Microsoft launches Muse AI for game creation and classic game preservation
- Luma AI unveils Art in Motion with Ray2 transforming paintings into moving masterpieces effortlessly
- Google AI introduces co-scientist system to help scientists generate hypotheses and strategies
- Grok to launch as standalone app for MacOS and Windows expanding accessibility and usability
- Convergence AI launches Proxy 1.0 for faster cheaper web automation
- ByteDance introduces Phantom for identity-preserving and subject-consistent AI video generation
- LM Studio 0.3.10 adds Speculative Decoding for 2x faster inferencing with no quality loss

Thu, Feb 20
- Google Gemini now allows all users to upload docs for summaries insights and workflow efficiency
- OpenAI to launch GPT-4.5 next week and GPT-5 in May with chain-of-thought reasoning advancements
- LTX Studio exits beta with new tools for video creation and upgraded creative workflows
- Pika introduces Pikaswaps for innovative AI video editing
- Alibaba unveils WanX 2.1 open-source AI model for next-gen video generation coming soon
- Midjourney enhances website with improved folder system and seamless image organization features
- Microsoft Research releases BioEmu-1 AI for ultra-fast protein structure prediction with lab-level accuracy
- Spotify partners with ElevenLabs for AI-narrated audiobooks
- OpenAI surpasses 400M weekly active users marking 33% growth in under three months
- Elon Musk announces Grok voice mode with adjustable personalities for personalized AI interaction
- Figure launches Helix, an AI humanoid robot for home automation
- xAI releases Grok 3 for free boasting world’s smartest AI until server capacity lasts

Fri, Feb 21
- Fal AI now offers Google DeepMind's Veo 2 video generation model
- 1X Tech introduces NEO Gamma, advancing home robotics technology
- Georgia Tech trains humanoid robots using Meta’s Project Aria glasses data
- Apple releases iOS 18.4 beta with expanded Apple Intelligence features
- Apple Intelligence to boost Vision Pro with visionOS 2.4 update adding AI tools and remote management
- Google launches Veo 2 AI video generator on Freepik with 4K support and free access for early users
- OpenAI expands Operator AI agent to ChatGPT Pro users in 8 countries with global rollout planned
- DeepSeek to open-source 5 AI code repositories next week challenging industry norms for transparency

Sat, Feb 22
- Kimi AI introduces Muon, a scalable optimizer for efficient LLM training
- Hailuo makes I2V-01-Director image-to-video model available for all users
- Luma AI adds Upscale to its Ray2 API for developers

Sun, Feb 23
- Apple prepares Google Gemini integration for Apple Intelligence features