That’s a wrap on this week’s AI updates. Follow us @MantleX_AI to stay ahead of the curve and never miss the biggest developments in the space.

Hugging Face launches AI Sheets. An open-source, no-code spreadsheet tool for building and transforming datasets with AI. Supports model selection from the Hugging Face Hub or local deployments, enabling quick experiments and “vibe testing” without writing code.

Cursor releases AI coding CLI Beta. Brings GPT-5, Claude, and Gemini straight into the terminal for real-time AI coding help. Features include inline edits, customisable rules, and smooth integration with IDEs. Early users say it makes development faster and more fluid.

Tsinghua University team breaks a 41-year-old record! Researchers developed a new shortest-path algorithm that removes the “sorting bottleneck” present in existing methods. It earned the Best Paper Award at STOC 2025 and could reshape route optimisation across industries.

Claude Code adds automated security reviews. Developers can now run "/security-review" in the terminal or integrate checks via GitHub Actions to catch vulnerabilities like SQL injection, XSS, and insecure authentication. The feature has already flagged real-world flaws, helping

More big moves in AI this week (Part 2): - Claude Code gets automated security reviews - Tsinghua breaks 41-year-old shortest-path record - Cursor launches AI coding CLI - Hugging Face ships AI Sheets Let's break it down ⬇️

Follow @MantleX_AI to stay updated on the latest AI developments and breakthroughs. We have Part 2 coming tomorrow, you don't want to miss it! If you found this thread valuable, consider leaving a like/repost of the post below.

xAI makes Grok 4 freely available worldwide with Auto and Expert modes accessible to all users and significantly increased usage limits. Only the most powerful variant, Grok 4 Heavy, remains subscription-only. The timing appears strategic following GPT-5's reception issues.

OpenAI's GPT-5 rollout faced significant user pushback, with complaints about reduced capability and disrupted workflows. The company reinstated GPT-4o access for Plus users and acknowledged that removing legacy models was a misstep, highlighting the tension between technical

Google's new Guided Learning mode transforms Gemini into an interactive tutor, delivering step-by-step explanations, adaptive questions, and multimodal teaching support including quizzes, diagrams, and videos. A free one-year AI Pro subscription is available for students in

ElevenLabs introduces Eleven Music for AI-powered full-track generation from natural language prompts. The tool offers granular control over genre, style, structure, lyrics, and vocals, with real-time editing capabilities and variation generation—opening professional-grade music

Anthropic releases Claude Opus 4, delivering best-in-class coding performance with 72.5% on SWE-bench compared to GPT-4.1's 54.6%. The model excels at reasoning across large codebases and can work autonomously for up to seven hours. Opus 4.1 further improves performance to 74.5%

DeepMind's Genie 3 generates fully interactive 3D worlds in real time at 720p and 24 FPS. Users can explore and modify scenes through text prompts—changing weather, inserting objects—with visual consistency maintained for several minutes. A significant leap from Genie 2's

Another packed week in AI (Part 1): - DeepMind launches Genie 3 - Claude Opus 4 tops coding benchmarks - ElevenLabs ships AI music studio - Google Gemini adds Guided Learning - GPT-5 faces backlash - Grok 4 goes free worldwide Here’s everything you need to know ⬇️

The AI landscape is evolving at breakneck speed, and staying informed isn't optional anymore. If you found this breakdown valuable, hit follow and like this thread so you don't miss the next major developments shaping our technological future.

🔟 Scientific Discovery Platform GPT-5 establishes the foundation for AI systems capable of autonomous hypothesis generation and testing. This positions AI as a potential catalyst for breakthrough discoveries across scientific disciplines.

9️⃣ Efficient Architecture New algorithmic advances enable powerful capabilities in smaller, more efficient models. This allows for local deployment, reducing costs while enhancing privacy protection.

8️⃣ Improved Accuracy & Reliability Enhanced training protocols significantly reduce hallucinations and improve factual alignment. This increased reliability makes the model suitable for high-stakes applications in education and healthcare.

7️⃣ Seamless Integration GPT-5 integrates proactively across digital platforms—email, calendar, documents—while processing multiple input types. This transforms it from a reactive tool into a comprehensive digital assistant.

6️⃣ Advanced Coding Capabilities The model demonstrates substantial improvements in code generation, debugging, and logical reasoning.These capabilities support complex automation and problem-solving that previously required human expertise.