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.
LightOn released ColGrep-Agent-2B, an Apache-2.0 local code-search agent. Delegating search cut Claude Code API tokens 34% in its tests, at a time cost.
CoreWeave Forge opens Model Distillation in private preview: capture traffic, train a smaller open-weight model, compare and roll out. Billing and caveats.
Ex-DeepMind researcher David Silver pitches an RL superlearner as another road to superintelligence. What Ineffable says and what is verified.
DigitalOcean launched Signals, Simulations and Insights in public preview, plus Agent Data Services in private preview, to find and fix silent agent failures.
Epic Games reportedly delayed its Disney shooter and told staff "no more hand-written code." What Bloomberg reported, what Epic said, and what is unconfirmed.
FERC rejected most of PJM's large-load plan on Oct 9, while Entergy reportedly warned Amazon of grid cuts. What is verified, what is claimed, what it means.
Microsoft researchers say long-running agents get stuck in idea basins. Fork-and-Flush clears chat context and beat best-of-4 by 44.4%. What to copy.
Z.ai unveiled GLM-Legal, a legal agent model on GLM-5.3-Flash with a 19.2% Harvey LAB all-pass rate. Access is gated. What is claimed vs verified.
Andrej Karpathy says about 5,000 people have internal frontier AI access while 75% of the world barely used LLMs. What he said and what is unverified.
Lemo-Opuscar is an MIT Claude Code skill with 43 film styles that renders short films as code with Opus 5.5. Setup, cost, limits, comparison.
Reuters says McDonald's uses machine learning to recommend prices at 14,000 restaurants. A class action followed; McDonald's says AI does not set prices.
Meta launched Muse Wonder and Muse Wonder Full for coding and agents. Wonder costs $4 input and $20 output per million tokens; Full needs Muse Code.
