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.
Inkling-Small: 12B active MoE open weights that match Inkling on coding and reasoning at lower cost. Benchmarks, Tinker, trade-offs vs factuality.
EvoLib: test-time learning via consolidation + dynamic weighting of skills and insights. MIT code, arXiv 2605.14477, vs memory archives and SkillOpt.
Microsoft’s generative-ai-for-beginners is a free 21-lesson course (115K+ stars). 2026 updates: Foundry Models, Responses API path, GitHub Models retirement.
Paul Graham says LLMs excel at math because answers are clearly right or wrong—not because math is easier. What that means for training, writing, and product evals.
Qwen’s Aug 3 video calls Qwen3.8-Max a new bar for coding and cowork. See access paths, discounts, and why open weights remain the real ask.
Tencent open-sourced Agent Memory v2.0.0: Chat Memory, Skills, LLM Wiki, and CodeGraph with ACL loadouts for OpenClaw, Hermes, Claude Code, and more.
Anthropic’s Thariq says AI already shows Jevons paradox in math: more activity, easier understanding, higher abstraction — and rising demand for people who know math. Chess parallel decoded.
Musk welcomed the world to the singularity after Astra’s math results. See what the term means, why leaders disagree, and what actually changes.
Anatomy Atelier: 3D organs built with GPT Image 2.0, TripoAI, and Codex, optimized from 900MB to 28.6MB. Live demo, GitHub repo, and how it works.
FFmpeg says Cursor gave several of its developers free credits for coding and code review. What was announced, the reaction, and why it matters.
Cursor replaced dollar estimates with tokens, users objected, and the company reversed it. Here is what AI coding cost transparency should show.
Fable-OS puts Claude behind 64 kernel tools in QEMU. See what the audio-driver demo proves—and the Ring 0, TLS, hardware, and license caveats.
