Expert profile
Founder & AI Product Leader
Yash Thakker is a Generative AI expert with over 12 years of experience in product leadership and technical strategy. As the founder of explainx.ai, he has taught over 300,000 learners and built AI platforms serving millions of users globally. He specializes in Agentic AI, Multimodal RAG, and the intersection of LLMs with consumer hardware.
Google Research combines four agentic methods to keep AI video consistent over minutes. How Co-Director, CANVAS, A²RD and VQQA fight drift and what builders can reuse.
Anthropic says Claude cut claude.ai load times 3.1x on average in 14 days, with 150+ parallel threads and zero incidents. The method, numbers and playbook to copy.
Menlo Research open-sourced the RL training code behind Asimov 1's walking policy — Isaac Lab, PPO, and Adversarial Motion Priors, BSD-3 licensed.
Meta Connect named GitHub as a Muse connector; the Model API cookbook shows Muse Spark triaging issues, reviewing PRs, and opening fix PRs after approval.
The Verge says Muse can export hundreds of MB of its own files. Meta and Nat Friedman call it intended: your own Linux VM. What was exposed and what still matters.
Microsoft's September 25, 2026 Copilot overhaul adds Home, Code, and Autopilot — plus full Word, Excel, and PowerPoint built into Copilot.
Elon Musk is "cautiously optimistic" SpaceX will match Fable and GPT-6 in 2 to 3 months. What he claimed, what Grok 4.8 shows, and how builders should read it.
Odyssey released a playable Agora-2 preview: up to 20 humans or AI agents in one real-time generated world with shared state and no traditional game engine.
Leaks point to a $500 ChatGPT Pro Max plan with fastest Work and Codex while the $200 Pro tier stays paused for new sign-ups amid Astra demand.
The Information reports Google, OpenAI, and Anthropic may launch SAFA, a private standards authority for frontier models, after a public-private safety partnership stalled in Washington.
OpenClaw removed about 400K lines of its own tests with little coverage change using a test-audit agent skill. The four-question gate and how to copy it.
Perplexity replaced its search engine with Photon, a Rust retrieval service behind Fast Search: 95% of results in 230 ms and 68% cheaper agent tasks, with trade-offs.
