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.
Rigel, a 2.3B model, reportedly matches Llama-3.2-3B using under 1% of comparable pretraining compute. What the claim says, and what to verify.
Stripe upgraded 7.8M hosted checkout pages to cut AI agent token use 42%. What changed, and why it matters for agents navigating checkout.
Anthropic's Thariq argues smarter models should mean more time understanding users, not more shipped features. The argument and pushback.
Unreal Labs' new harness claims 40% cost savings over Codex via async tool calls. Benchmarks, architecture, and the Epic name confusion.
Z.ai open-sourced ZCode after its default config was found uploading user codebases to Alibaba Cloud. What happened, and what to check.
a16z's Horowitz Andreessen Academy: $35M, tuition-free, ages 16-22, AI and startup focus, co-ops at Anthropic/Meta/Stripe. Founding Class fall 2027.
A viral tier list ranked Oh My Pi S-tier and Claude Code/Codex B-tier — with Hermes missing entirely. Here's what holds up and what doesn't.
Grep.ai's AgentRun lets an agent do a job once, then write itself out of the loop as a workflow — cutting AML alert cost from $2.89 to $0.25.
Amazon cut off Meta's Muse agent from shopping on Amazon.com, citing undisclosed access, no self-identification, and credential capture.
Andrew Ng's Batch 371 letter: AI fear is overhyped, Hugging Face swarm headlines misread parallelism, cyber risk is real but fixable — keep building.
Cloudflare Worker Previews: npx wrangler preview gives every git branch its own Worker, URL, Durable Objects, and logs. Ships in Wrangler 4.135.0.
SpaceXAI shipped Grok 4.7 on Sept 21 at the same $2/$6 as 4.6. It leads EEBench and Harvey Legal, trails Fable 5.1 on CursorBench.
